{
 "titre": "Le même modèle, des outils différents : le score selon l'agent",
 "url": "https://quelleia.com/outils/harnais/",
 "edition": "2026-W41",
 "date": "2026-10-05",
 "bilan": {
  "mesures": 393,
  "ecritures": 52,
  "harnais": 44,
  "doublons": 70,
  "retirees": 7,
  "renommees": 3,
  "couples_harnais": 30,
  "couples_pas": 0,
  "modeles": 30,
  "outils_grand_public": 15,
  "inconnues": 0
 },
 "types": [
  {
   "id": "grand_public",
   "nom": "outil grand public",
   "definition": "un outil qu'on installe et qu'on utilise soi-même"
  },
  {
   "id": "cadriciel",
   "nom": "bibliothèque pour développeurs",
   "definition": "une bibliothèque de code avec laquelle on programme son propre agent"
  },
  {
   "id": "banc",
   "nom": "outil de banc d'essai",
   "definition": "l'agent de référence d'un banc d'essai, fait pour mesurer des modèles"
  },
  {
   "id": "recherche",
   "nom": "agent de recherche",
   "definition": "un agent publié par un laboratoire ou une université"
  },
  {
   "id": "autre",
   "nom": "autre agent",
   "definition": "un agent d'une société ou d'un particulier, soumis au classement"
  },
  {
   "id": "non_precise",
   "nom": "agent non précisé",
   "definition": "la source ne dit pas quel agent pilotait le modèle"
  },
  {
   "id": "inconnu",
   "nom": "écriture non rangée",
   "definition": "un nom que le registre ne connaît pas encore"
  }
 ],
 "limites": [
  "Un seul benchmark mesure un même modèle dans plusieurs outils : Terminal-Bench 2.0. Un outil fort en terminal peut l'être moins ailleurs.",
  "OSWorld n'est pas repris ici. Sur OSWorld, les lignes d'un même modèle diffèrent par le nombre d'étapes autorisées (15, 50 ou 100) avec le même outil : elles ne comparent donc pas des outils.",
  "Les modèles mesurés ont souvent plusieurs mois. Les outils changent chaque semaine : une mesure ne vaut que pour la version de l'outil soumise ce jour-là.",
  "Beaucoup d'agents mesurés viennent de laboratoires ou ont été faits pour une démonstration, en visant le classement. Les outils qu'on installe soi-même sont en jaune.",
  "Chaque équipe lance sa propre mesure, puis la soumet. Le classement Terminal-Bench 2.0 exige au moins cinq essais par tâche, sans toucher aux délais ni aux ressources, et fait relire les soumissions. Il a déjà retiré des agents qui trichaient.",
  "Le classement Terminal-Bench date chaque ligne du jour de sortie du modèle. Le jour de la mesure n'est pas publié.",
  "Le plus souvent, il n'existe qu'une mesure par outil et par modèle. Deux soumissions peuvent différer de quelques points sans que l'outil ait changé."
 ],
 "note_date": "Le classement Terminal-Bench date chaque ligne du jour de sortie du modèle et ne publie pas la date de chaque mesure.",
 "sources": [
  {
   "id": "terminal_bench",
   "nom": "Terminal-Bench 2.0",
   "url": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
   "texte": "Classement tenu par l'équipe de Terminal-Bench (Stanford et Laude Institute). Les soumissions sont faites par les équipes elles-mêmes, vérifiées par un robot puis relues par les mainteneurs. Données sous licence Apache 2.0 (dépôt des soumissions).",
   "conditions": "https://huggingface.co/datasets/harborframework/terminal-bench-2-leaderboard",
   "regles": "https://www.tbench.ai/news/leaderboard-integrity-update"
  }
 ],
 "note_epoch": "Ces mesures arrivent par les fichiers « external » d'Epoch AI, qui reprennent les classements de leurs auteurs. Epoch précise que ces données gardent la licence de leur source : on vérifie donc chez chaque classement ce qu'on a le droit d'en reprendre.",
 "source_des_mesures": {
  "nom": "Epoch AI",
  "url": "https://epoch.ai/benchmarks"
 },
 "harnais": [
  {
   "id": "ante",
   "nom": "Ante",
   "type": "autre",
   "editeur": "Antigma Labs",
   "lien": "https://antigma.ai",
   "lien_source": "https://huggingface.co/datasets/harborframework/terminal-bench-2-leaderboard/blob/main/submissions/terminal-bench/2.0/Ante__Gemini-3.1-Pro-Preview/metadata.yaml",
   "fiche": "https://quelleia.com/outils/ante/",
   "ecritures": [
    "Ante"
   ],
   "note": null
  },
  {
   "id": "camel-ai",
   "nom": "CAMEL-AI",
   "type": "recherche",
   "editeur": "CAMEL-AI",
   "lien": null,
   "lien_source": null,
   "fiche": "https://quelleia.com/outils/camel-ai/",
   "ecritures": [
    "CAMEL-AI"
   ],
   "note": null
  },
  {
   "id": "capy",
   "nom": "Capy",
   "type": "autre",
   "editeur": "Capy",
   "lien": "https://capy.ai",
   "lien_source": "https://huggingface.co/datasets/harborframework/terminal-bench-2-leaderboard/blob/main/submissions/terminal-bench/2.0/Capy__GPT-5.5/metadata.yaml",
   "fiche": "https://quelleia.com/outils/capy/",
   "ecritures": [
    "Capy"
   ],
   "note": null
  },
  {
   "id": "cchuter",
   "nom": "cchuter",
   "type": "autre",
   "editeur": "teamblobfish.com",
   "lien": "https://github.com/cchuter/blobfish",
   "lien_source": "https://huggingface.co/datasets/harborframework/terminal-bench-2-leaderboard/blob/main/submissions/terminal-bench/2.0/cchuter__minimax-m2.5/metadata.yaml",
   "fiche": "https://quelleia.com/outils/cchuter/",
   "ecritures": [
    "cchuter"
   ],
   "note": null
  },
  {
   "id": "chisel",
   "nom": "chisel",
   "type": "autre",
   "editeur": null,
   "lien": null,
   "lien_source": null,
   "fiche": "https://quelleia.com/outils/chisel/",
   "ecritures": [
    "chisel"
   ],
   "note": "Agent utilisé sur le classement FrontierCode de Cognition. La page du classement ne le décrit pas, et son éditeur n'a pas été vérifié."
  },
  {
   "id": "claude-code",
   "nom": "Claude Code",
   "type": "grand_public",
   "editeur": "Anthropic",
   "lien": "https://claude.com/product/claude-code",
   "lien_source": "https://huggingface.co/datasets/harborframework/terminal-bench-2-leaderboard/blob/main/submissions/terminal-bench/2.0/ClaudeCode__GLM-4.7/metadata.yaml",
   "fiche": "https://quelleia.com/outils/claude-code/",
   "ecritures": [
    "Claude Code",
    "claude-code",
    "claude_code"
   ],
   "note": null
  },
  {
   "id": "clnkr",
   "nom": "clnkr",
   "type": "autre",
   "editeur": "clnkr",
   "lien": "https://github.com/clnkr-ai/clnkr",
   "lien_source": "https://huggingface.co/datasets/harborframework/terminal-bench-2-leaderboard/blob/main/submissions/terminal-bench/2.0/clnkr__GPT-5.5/metadata.yaml",
   "fiche": "https://quelleia.com/outils/clnkr/",
   "ecritures": [
    "clnkr"
   ],
   "note": null
  },
  {
   "id": "codebrain-1-5",
   "nom": "CodeBrain-1.5",
   "type": "autre",
   "editeur": "Feeling AI",
   "lien": "https://feeling-ai.com",
   "lien_source": "https://huggingface.co/datasets/harborframework/terminal-bench-2-leaderboard/blob/main/submissions/terminal-bench/2.0/CodeBrain-1.5__GPT-5.3-Codex/metadata.yaml",
   "fiche": "https://quelleia.com/outils/codebrain-1-5/",
   "ecritures": [
    "CodeBrain-1.5"
   ],
   "note": "Version 1.5 de CodeBrain, comptée à part de CodeBrain-1. Le 19 mai 2026, le classement a affiché sous CodeBrain-1.5 la ligne Gemini 3 Pro de CodeBrain-1."
  },
  {
   "id": "codelia",
   "nom": "Codelia",
   "type": "autre",
   "editeur": "kousw",
   "lien": "https://github.com/kousw/codelia",
   "lien_source": "https://huggingface.co/datasets/harborframework/terminal-bench-2-leaderboard/blob/main/submissions/terminal-bench/2.0/Codelia__GPT-5.3-Codex/metadata.yaml",
   "fiche": "https://quelleia.com/outils/codelia/",
   "ecritures": [
    "Codelia"
   ],
   "note": null
  },
  {
   "id": "codex-cli",
   "nom": "Codex CLI",
   "type": "grand_public",
   "editeur": "OpenAI",
   "lien": "https://openai.com/codex/",
   "lien_source": "https://raw.githubusercontent.com/harbor-framework/terminal-bench/main/leaderboard/src/leaderboard/display-names.json",
   "fiche": "https://quelleia.com/outils/codex/",
   "ecritures": [
    "Codex CLI",
    "codex",
    "codex_cli",
    "Codex"
   ],
   "note": "« Codex » et « Codex CLI » désignent le même agent d'OpenAI (même éditeur, même lien au classement) ; ia/schema.py les unifie déjà. Sur GPT-5.5, deux soumissions distinctes existent (82,2 % et 82,0 %)."
  },
  {
   "id": "crux",
   "nom": "Crux",
   "type": "autre",
   "editeur": "Roam",
   "lien": "https://github.com/0xDarkMatter/crux-agent",
   "lien_source": "https://huggingface.co/datasets/harborframework/terminal-bench-2-leaderboard/blob/main/submissions/terminal-bench/2.0/Crux__Claude-Opus-4.6/metadata.yaml",
   "fiche": "https://quelleia.com/outils/crux/",
   "ecritures": [
    "Crux"
   ],
   "note": null
  },
  {
   "id": "cursor-cli",
   "nom": "Cursor CLI",
   "type": "grand_public",
   "editeur": "Anysphere",
   "lien": null,
   "lien_source": null,
   "fiche": "https://quelleia.com/outils/cursor/",
   "ecritures": [
    "Cursor CLI",
    "cursor-cli",
    "Cursor"
   ],
   "note": "« Cursor CLI » (PostTrainBench) et « cursor-cli » (FrontierCode) : même nom à la casse et au tiret près ; c'est l'agent en ligne de commande de Cursor."
  },
  {
   "id": "dakou-agent",
   "nom": "Dakou Agent",
   "type": "autre",
   "editeur": "iflow",
   "lien": "https://dakou.iflow.cn/",
   "lien_source": "https://huggingface.co/datasets/harborframework/terminal-bench-2-leaderboard/blob/main/submissions/terminal-bench/2.0/dakou__qwen3-coder-480b/metadata.yaml",
   "fiche": "https://quelleia.com/outils/dakou-agent/",
   "ecritures": [
    "Dakou Agent"
   ],
   "note": null
  },
  {
   "id": "deep-agents",
   "nom": "Deep Agents",
   "type": "cadriciel",
   "editeur": "LangChain",
   "lien": "https://github.com/langchain-ai/deepagents",
   "lien_source": "https://huggingface.co/datasets/harborframework/terminal-bench-2-leaderboard/blob/main/submissions/terminal-bench/2.0/Deep-Agents__GPT-5.2-Codex/metadata.yaml",
   "fiche": "https://quelleia.com/outils/deep-agents/",
   "ecritures": [
    "Deep Agents",
    "deepagents"
   ],
   "note": null
  },
  {
   "id": "devin",
   "nom": "Devin",
   "type": "grand_public",
   "editeur": "Cognition",
   "lien": null,
   "lien_source": null,
   "fiche": "https://quelleia.com/outils/devin/",
   "ecritures": [
    "devin",
    "Devin"
   ],
   "note": "L'agent en ligne Devin de Cognition, qui tient aussi le classement FrontierCode ; distinct de Devin Desktop (ex-Windsurf)."
  },
  {
   "id": "droid",
   "nom": "Droid (Factory)",
   "type": "grand_public",
   "editeur": "Factory",
   "lien": "https://factory.ai/",
   "lien_source": "https://huggingface.co/datasets/harborframework/terminal-bench-2-leaderboard/blob/main/submissions/terminal-bench/2.0/Droid__Claude-Opus-4.6/metadata.yaml",
   "fiche": "https://quelleia.com/outils/factory-droid/",
   "ecritures": [
    "Droid"
   ],
   "note": null
  },
  {
   "id": "gemini-cli",
   "nom": "Gemini CLI",
   "type": "grand_public",
   "editeur": "Google",
   "lien": "https://github.com/google-gemini/gemini-cli",
   "lien_source": "https://huggingface.co/datasets/harborframework/terminal-bench-2-leaderboard/blob/main/submissions/terminal-bench/2.0/Gemini_CLI__Gemini-3.1-Pro-Preview/metadata.yaml",
   "fiche": "https://quelleia.com/outils/gemini-cli/",
   "ecritures": [
    "Gemini CLI",
    "gemini_cli",
    "gemini-cli"
   ],
   "note": null
  },
  {
   "id": "goose",
   "nom": "goose",
   "type": "grand_public",
   "editeur": "Block",
   "lien": null,
   "lien_source": null,
   "fiche": "https://quelleia.com/outils/goose/",
   "ecritures": [
    "Goose"
   ],
   "note": null
  },
  {
   "id": "grok-build",
   "nom": "Grok Build",
   "type": "grand_public",
   "editeur": "xAI",
   "lien": "https://x.ai/build",
   "lien_source": "https://raw.githubusercontent.com/harbor-framework/terminal-bench/main/leaderboard/src/leaderboard/display-names.json",
   "fiche": "https://quelleia.com/outils/grok-build/",
   "ecritures": [
    "grok-build",
    "Grok Build"
   ],
   "note": null
  },
  {
   "id": "grok-cli",
   "nom": "Grok CLI",
   "type": "grand_public",
   "editeur": "Superagent",
   "lien": "https://github.com/superagent-ai/grok-cli",
   "lien_source": "https://huggingface.co/datasets/harborframework/terminal-bench-2-leaderboard/blob/main/submissions/terminal-bench/2.0/grok-cli__grok-4.20-0309-reasoning/metadata.yaml",
   "fiche": "https://quelleia.com/outils/grok-cli/",
   "ecritures": [
    "Grok CLI",
    "grok-cli"
   ],
   "note": "Projet open source de Superagent, distinct de Grok Build (xAI). « Grok CLI » (éditeur déclaré Superagent) et « grok-cli » (Vibe Kit) : même nom à la casse près, même modèle (Grok 4.20) et même score à trois décimales (57,3 %) ; Epoch a saisi deux fois la même ligne."
  },
  {
   "id": "harness-agent",
   "nom": "Harness Agent",
   "type": "autre",
   "editeur": "lazyFrogLOL",
   "lien": "https://github.com/lazyFrogLOL/Harness_Engineering",
   "lien_source": "https://huggingface.co/datasets/harborframework/terminal-bench-2-leaderboard/blob/main/submissions/terminal-bench/2.0/harness-agent__minimax-m2-7-highspeed/metadata.yaml",
   "fiche": "https://quelleia.com/outils/harness-agent/",
   "ecritures": [
    "Harness Agent"
   ],
   "note": null
  },
  {
   "id": "hookele",
   "nom": "hookele",
   "type": "autre",
   "editeur": "Dmitry Barakhov",
   "lien": "https://github.com/sady4850/hookele_coding_agent",
   "lien_source": "https://huggingface.co/datasets/harborframework/terminal-bench-2-leaderboard/blob/main/submissions/terminal-bench/2.0/hookele__gpt5.1-codex-mini/metadata.yaml",
   "fiche": "https://quelleia.com/outils/hookele/",
   "ecritures": [
    "hookele"
   ],
   "note": null
  },
  {
   "id": "ii-agent",
   "nom": "II-Agent",
   "type": "autre",
   "editeur": "Intelligent Internet",
   "lien": null,
   "lien_source": null,
   "fiche": "https://quelleia.com/outils/ii-agent/",
   "ecritures": [
    "II-Agent"
   ],
   "note": null
  },
  {
   "id": "indusagi",
   "nom": "IndusAGI Coding Agent",
   "type": "autre",
   "editeur": "Varun Israni (SoloVpx)",
   "lien": "https://github.com/varun-israni/indusagi-coding-agent",
   "lien_source": "https://huggingface.co/datasets/harborframework/terminal-bench-2-leaderboard/blob/main/submissions/terminal-bench/2.0/IndusAGICodingAgent__MiniMax-M2.7/metadata.yaml",
   "fiche": "https://quelleia.com/outils/indusagi/",
   "ecritures": [
    "IndusAGI Coding Agent"
   ],
   "note": null
  },
  {
   "id": "junie-cli",
   "nom": "Junie CLI",
   "type": "grand_public",
   "editeur": "JetBrains",
   "lien": "https://junie.jetbrains.com",
   "lien_source": "https://huggingface.co/datasets/harborframework/terminal-bench-2-leaderboard/blob/main/submissions/terminal-bench/2.0/Junie_CLI__Gemini-3-Flash-Preview-Gemini-3.1-Pro-Preview-Claude-Opus-4.6-GPT-5.3-Codex/metadata.yaml",
   "fiche": "https://quelleia.com/outils/junie/",
   "ecritures": [
    "Junie CLI"
   ],
   "note": "La soumission déclare quatre modèles (Gemini 3 Flash, Gemini 3.1 Pro, Claude Opus 4.6, GPT-5.3 Codex) : Junie choisit le modèle selon la tâche."
  },
  {
   "id": "kimi-code",
   "nom": "Kimi Code CLI",
   "type": "grand_public",
   "editeur": "Moonshot AI",
   "lien": null,
   "lien_source": null,
   "fiche": "https://quelleia.com/outils/kimi-code/",
   "ecritures": [
    "kimi_code",
    "Kimi Code",
    "Kimi Code CLI"
   ],
   "note": null
  },
  {
   "id": "letta-code",
   "nom": "Letta Code",
   "type": "grand_public",
   "editeur": "Letta",
   "lien": null,
   "lien_source": null,
   "fiche": "https://quelleia.com/outils/letta-code/",
   "ecritures": [
    "Letta Code"
   ],
   "note": null
  },
  {
   "id": "little-coder",
   "nom": "little-coder",
   "type": "autre",
   "editeur": "Itay Inbar",
   "lien": "https://github.com/itayinbarr/little-coder",
   "lien_source": "https://huggingface.co/datasets/harborframework/terminal-bench-2-leaderboard/blob/main/submissions/terminal-bench/2.0/little-coder__qwen3.6-35b-a3b/metadata.yaml",
   "fiche": "https://quelleia.com/outils/little-coder/",
   "ecritures": [
    "little-coder"
   ],
   "note": null
  },
  {
   "id": "maya-v2",
   "nom": "MAYA-V2",
   "type": "autre",
   "editeur": "ADYA",
   "lien": "https://adya.ai/maya",
   "lien_source": "https://huggingface.co/datasets/harborframework/terminal-bench-2-leaderboard/blob/main/submissions/terminal-bench/2.0/MAYA__Claude-4.6-opus/metadata.yaml",
   "fiche": "https://quelleia.com/outils/maya-v2/",
   "ecritures": [
    "MAYA-V2"
   ],
   "note": "Deuxième version de MAYA, comptée à part. Le classement affiche sous MAYA-V2 la ligne Claude Sonnet 4.5 d'abord publiée sous MAYA."
  },
  {
   "id": "meta-harness",
   "nom": "Meta-Harness",
   "type": "recherche",
   "editeur": "Stanford IRIS",
   "lien": "https://github.com/stanford-iris-lab/meta-harness-tbench2-artifact",
   "lien_source": "https://huggingface.co/datasets/harborframework/terminal-bench-2-leaderboard/blob/main/submissions/terminal-bench/2.0/Meta-Harness__Claude-Opus-4.6/metadata.yaml",
   "fiche": "https://quelleia.com/outils/meta-harness/",
   "ecritures": [
    "Meta-Harness"
   ],
   "note": null
  },
  {
   "id": "mini-swe-agent",
   "nom": "mini-SWE-agent",
   "type": "banc",
   "editeur": "Princeton",
   "lien": null,
   "lien_source": null,
   "fiche": "https://quelleia.com/outils/mini-swe-agent/",
   "ecritures": [
    "mini-SWE-agent",
    "mini-swe-agent",
    "Mini-SWE-Agent"
   ],
   "note": "Éditeur tel qu'Epoch le recopie du classement Terminal-Bench (colonne « Agent Org »)."
  },
  {
   "id": "mux",
   "nom": "Xum (ex-Mux)",
   "type": "grand_public",
   "editeur": "Coder",
   "lien": "https://github.com/coder/mux",
   "lien_source": "https://huggingface.co/datasets/harborframework/terminal-bench-2-leaderboard/blob/main/submissions/terminal-bench/2.0/Mux__Claude-Opus-4.6/metadata.yaml",
   "fiche": "https://quelleia.com/outils/mux/",
   "ecritures": [
    "Mux"
   ],
   "note": null
  },
  {
   "id": "nexau-ahe",
   "nom": "NexAU-AHE",
   "type": "autre",
   "editeur": "china-qijizhifeng",
   "lien": "https://github.com/china-qijizhifeng/agentic-harness-engineering",
   "lien_source": "https://huggingface.co/datasets/harborframework/terminal-bench-2-leaderboard/blob/main/submissions/terminal-bench/2.0/NexAU-AHE__gpt-5.5/metadata.yaml",
   "fiche": "https://quelleia.com/outils/nexau-ahe/",
   "ecritures": [
    "NexAU-AHE"
   ],
   "note": null
  },
  {
   "id": "opencode",
   "nom": "OpenCode",
   "type": "grand_public",
   "editeur": "Anomaly Innovations",
   "lien": "https://opencode.ai",
   "lien_source": "https://huggingface.co/datasets/harborframework/terminal-bench-2-leaderboard/blob/main/submissions/terminal-bench/2.0/OpenCode__Claude-Opus-4.5/metadata.yaml",
   "fiche": "https://quelleia.com/outils/opencode/",
   "ecritures": [
    "OpenCode",
    "opencode"
   ],
   "note": "« OpenCode » (PostTrainBench, Terminal-Bench) et « opencode » (WeirdML) : même nom à la casse près, un seul outil de ce nom au registre."
  },
  {
   "id": "openhands",
   "nom": "OpenHands",
   "type": "grand_public",
   "editeur": "OpenHands",
   "lien": null,
   "lien_source": null,
   "fiche": "https://quelleia.com/outils/openhands/",
   "ecritures": [
    "OpenHands"
   ],
   "note": null
  },
  {
   "id": "proximus",
   "nom": "proximus",
   "type": "banc",
   "editeur": null,
   "lien": null,
   "lien_source": null,
   "fiche": "https://quelleia.com/outils/proximus/",
   "ecritures": [
    "proximus"
   ],
   "note": "Seul agent mesuré sur le classement FrontierSWE, tenu par Proximal Labs (https://www.frontierswe.com/). La page du classement ne le décrit pas."
  },
  {
   "id": "sageagent",
   "nom": "SageAgent",
   "type": "autre",
   "editeur": "OpenSage",
   "lien": "https://github.com/opensage-agent",
   "lien_source": "https://huggingface.co/datasets/harborframework/terminal-bench-2-leaderboard/blob/main/submissions/terminal-bench/2.0/OpenSage__GPT-5.3-Codex/metadata.yaml",
   "fiche": "https://quelleia.com/outils/sageagent/",
   "ecritures": [
    "SageAgent"
   ],
   "note": null
  },
  {
   "id": "simplai-agent",
   "nom": "Simplai Agent",
   "type": "autre",
   "editeur": "SimplAI",
   "lien": "https://simplai.ai",
   "lien_source": "https://huggingface.co/datasets/harborframework/terminal-bench-2-leaderboard/blob/main/submissions/terminal-bench/2.0/Simplai-Agent__Claude-Sonnet-4.6/metadata.yaml",
   "fiche": "https://quelleia.com/outils/simplai-agent/",
   "ecritures": [
    "Simplai Agent"
   ],
   "note": null
  },
  {
   "id": "simple-codex",
   "nom": "Simple Codex",
   "type": "autre",
   "editeur": "OpenAI",
   "lien": "https://openai.com/codex",
   "lien_source": "https://huggingface.co/datasets/harborframework/terminal-bench-2-leaderboard/blob/main/submissions/terminal-bench/2.0/Simple-Codex__GPT-5.3-Codex/metadata.yaml",
   "fiche": "https://quelleia.com/outils/simple-codex/",
   "ecritures": [
    "Simple Codex"
   ],
   "note": "Soumis par OpenAI sous un autre nom que Codex CLI : le site le compte comme un agent distinct."
  },
  {
   "id": "spoox-o-m",
   "nom": "spoox-o-m",
   "type": "recherche",
   "editeur": "TUM",
   "lien": "https://github.com/plaume8/spoox",
   "lien_source": "https://huggingface.co/datasets/harborframework/terminal-bench-2-leaderboard/blob/main/submissions/terminal-bench/2.0/spoox-o-m__GPT-5.3-Codex/metadata.yaml",
   "fiche": "https://quelleia.com/outils/spoox-o-m/",
   "ecritures": [
    "spoox-o-m"
   ],
   "note": null
  },
  {
   "id": "terminus-2",
   "nom": "Terminus 2",
   "type": "banc",
   "editeur": "Terminal-Bench",
   "lien": "https://www.tbench.ai/terminus",
   "lien_source": "https://huggingface.co/datasets/harborframework/terminal-bench-2-leaderboard/blob/main/submissions/terminal-bench/2.0/Terminus2__Claude-Opus-4.6/metadata.yaml",
   "fiche": "https://quelleia.com/outils/terminus-2/",
   "ecritures": [
    "Terminus 2",
    "terminus-2",
    "Terminus2"
   ],
   "note": "L'agent de référence de Terminal-Bench ; Epoch le recopie avec l'éditeur « Stanford » ou « Terminal-Bench »."
  },
  {
   "id": "terminus-kira",
   "nom": "Terminus-KIRA",
   "type": "autre",
   "editeur": "KRAFTON AI",
   "lien": "https://www.krafton.ai/en/",
   "lien_source": "https://huggingface.co/datasets/harborframework/terminal-bench-2-leaderboard/blob/main/submissions/terminal-bench/2.0/Terminus-KIRA__Claude-Opus-4.6/metadata.yaml",
   "fiche": "https://quelleia.com/outils/terminus-kira/",
   "ecritures": [
    "Terminus-KIRA"
   ],
   "note": "Agent distinct de Terminus 2, l'agent de référence de Terminal-Bench."
  },
  {
   "id": "tongagents",
   "nom": "TongAgents",
   "type": "recherche",
   "editeur": "BIGAI",
   "lien": "https://www.bigai.ai/",
   "lien_source": "https://huggingface.co/datasets/harborframework/terminal-bench-2-leaderboard/blob/main/submissions/terminal-bench/2.0/Judy__Gemini-3.1-Pro-Preview/metadata.yaml",
   "fiche": "https://quelleia.com/outils/tongagents/",
   "ecritures": [
    "TongAgents",
    "Judy"
   ],
   "note": "Le classement affiche sous le nom TongAgents les soumissions « Judy » de BIGAI. Epoch écrit l'éditeur « BIGAI » ou « Bigai »."
  },
  {
   "id": "wozcode",
   "nom": "WOZCODE",
   "type": "autre",
   "editeur": "WOZCODE",
   "lien": "https://github.com/WithWoz/wozcode-plugin",
   "lien_source": "https://huggingface.co/datasets/harborframework/terminal-bench-2-leaderboard/blob/main/submissions/terminal-bench/2.0/WOZCODE__Claude-Opus-4-7/metadata.yaml",
   "fiche": "https://quelleia.com/outils/wozcode/",
   "ecritures": [
    "WOZCODE",
    "WozCode"
   ],
   "note": null
  }
 ],
 "couples": [
  {
   "id": "terminal_bench--claude-opus-4-6",
   "modele": "claude-opus-4-6",
   "modele_nom": "Claude Opus 4.6",
   "modele_url": "https://quelleia.com/modeles/claude-opus-4-6/",
   "benchmark": "terminal_bench",
   "benchmark_nom": "Terminal-Bench 2.0",
   "benchmark_url": "https://quelleia.com/benchmarks/terminal_bench/",
   "nature": "harnais",
   "harnais": 10,
   "ecart": 18.43,
   "barres": [
    {
     "harnais": "meta-harness",
     "nom": "Meta-Harness",
     "type": "recherche",
     "pas": null,
     "valeur": 76.4,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2026-02-05",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Meta-Harness"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/meta-harness/"
    },
    {
     "harnais": "capy",
     "nom": "Capy",
     "type": "autre",
     "pas": null,
     "valeur": 75.28,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2026-02-05",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Capy"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/capy/"
    },
    {
     "harnais": "terminus-kira",
     "nom": "Terminus-KIRA",
     "type": "autre",
     "pas": null,
     "valeur": 74.72,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2026-02-05",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Terminus-KIRA"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/terminus-kira/"
    },
    {
     "harnais": "maya-v2",
     "nom": "MAYA-V2",
     "type": "autre",
     "pas": null,
     "valeur": 72.08,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2026-02-05",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "MAYA-V2"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/maya-v2/"
    },
    {
     "harnais": "tongagents",
     "nom": "TongAgents",
     "type": "recherche",
     "pas": null,
     "valeur": 71.91,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2026-02-05",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "TongAgents"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/tongagents/"
    },
    {
     "harnais": "droid",
     "nom": "Droid (Factory)",
     "type": "grand_public",
     "pas": null,
     "valeur": 69.89,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2026-02-05",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Droid"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/factory-droid/"
    },
    {
     "harnais": "crux",
     "nom": "Crux",
     "type": "autre",
     "pas": null,
     "valeur": 66.87,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2026-02-05",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Crux"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/crux/"
    },
    {
     "harnais": "mux",
     "nom": "Xum (ex-Mux)",
     "type": "grand_public",
     "pas": null,
     "valeur": 66.52,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2026-02-05",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Mux"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/mux/"
    },
    {
     "harnais": "terminus-2",
     "nom": "Terminus 2",
     "type": "banc",
     "pas": null,
     "valeur": 62.92,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2026-02-05",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Terminus 2"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/terminus-2/"
    },
    {
     "harnais": "claude-code",
     "nom": "Claude Code",
     "type": "grand_public",
     "pas": null,
     "valeur": 57.98,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2026-02-05",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Claude Code"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/claude-code/"
    }
   ],
   "retirees": [
    {
     "harnais": "forgecode",
     "nom": "ForgeCode",
     "valeur": 79.8,
     "unite": "%",
     "sortie_modele": "2026-03-12",
     "constat": "Les trois lignes de ForgeCode figurent au classement Terminal-Bench 2.0 jusqu'au 13 mai 2026 et n'y sont plus à partir du 19 mai 2026. Le 19 avril 2026, l'équipe du classement écrivait avoir remis à zéro des essais où l'agent de ForgeCode récupérait la solution sur Internet.",
     "sources": [
      "https://web.archive.org/web/20260513024259/https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "https://web.archive.org/web/20260519111531/https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "https://www.tbench.ai/news/leaderboard-integrity-update"
     ]
    }
   ]
  },
  {
   "id": "terminal_bench--gpt-5-3-codex",
   "modele": "gpt-5-3-codex",
   "modele_nom": "GPT-5.3 Codex",
   "modele_url": "https://quelleia.com/modeles/gpt-5-3-codex/",
   "benchmark": "terminal_bench",
   "benchmark_nom": "Terminal-Bench 2.0",
   "benchmark_url": "https://quelleia.com/benchmarks/terminal_bench/",
   "nature": "harnais",
   "harnais": 9,
   "ecart": 13.71,
   "barres": [
    {
     "harnais": "sageagent",
     "nom": "SageAgent",
     "type": "autre",
     "pas": null,
     "valeur": 78.43,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2026-02-05",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "SageAgent"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/sageagent/"
    },
    {
     "harnais": "droid",
     "nom": "Droid (Factory)",
     "type": "grand_public",
     "pas": null,
     "valeur": 77.3,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2026-02-05",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Droid"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/factory-droid/"
    },
    {
     "harnais": "codebrain-1-5",
     "nom": "CodeBrain-1.5",
     "type": "autre",
     "pas": null,
     "valeur": 75.84,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2026-02-05",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "CodeBrain-1.5"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/codebrain-1-5/"
    },
    {
     "harnais": "codelia",
     "nom": "Codelia",
     "type": "autre",
     "pas": null,
     "valeur": 75.73,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2026-02-05",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Codelia"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/codelia/"
    },
    {
     "harnais": "simple-codex",
     "nom": "Simple Codex",
     "type": "autre",
     "pas": null,
     "valeur": 75.06,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2026-02-05",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Simple Codex"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/simple-codex/"
    },
    {
     "harnais": "mux",
     "nom": "Xum (ex-Mux)",
     "type": "grand_public",
     "pas": null,
     "valeur": 74.61,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2026-02-05",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Mux"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/mux/"
    },
    {
     "harnais": "spoox-o-m",
     "nom": "spoox-o-m",
     "type": "recherche",
     "pas": null,
     "valeur": 71.46,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2026-02-05",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "spoox-o-m"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/spoox-o-m/"
    },
    {
     "harnais": "indusagi",
     "nom": "IndusAGI Coding Agent",
     "type": "autre",
     "pas": null,
     "valeur": 69.14,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2026-02-05",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "IndusAGI Coding Agent"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/indusagi/"
    },
    {
     "harnais": "terminus-2",
     "nom": "Terminus 2",
     "type": "banc",
     "pas": null,
     "valeur": 64.72,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2026-02-05",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Terminus 2"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/terminus-2/"
    }
   ],
   "retirees": [
    {
     "harnais": "codebrain-1",
     "nom": "CodeBrain-1",
     "valeur": 70.3,
     "unite": "%",
     "sortie_modele": "2026-02-10",
     "constat": "Ligne présente jusqu'au 13 mai 2026 ; le 19 mai 2026, le classement montre à sa place CodeBrain-1.5 avec GPT-5.3 Codex, à 75,8 %.",
     "sources": [
      "https://web.archive.org/web/20260513024259/https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "https://web.archive.org/web/20260519111531/https://www.tbench.ai/leaderboard/terminal-bench/2.0"
     ]
    }
   ]
  },
  {
   "id": "terminal_bench--claude-opus-4-5",
   "modele": "claude-opus-4-5",
   "modele_nom": "Claude Opus 4.5",
   "modele_url": "https://quelleia.com/modeles/claude-opus-4-5/",
   "benchmark": "terminal_bench",
   "benchmark_nom": "Terminal-Bench 2.0",
   "benchmark_url": "https://quelleia.com/benchmarks/terminal_bench/",
   "nature": "harnais",
   "harnais": 8,
   "ecart": 11.46,
   "barres": [
    {
     "harnais": "droid",
     "nom": "Droid (Factory)",
     "type": "grand_public",
     "pas": null,
     "valeur": 63.15,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-11-24",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Droid"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/factory-droid/"
    },
    {
     "harnais": "letta-code",
     "nom": "Letta Code",
     "type": "grand_public",
     "pas": null,
     "valeur": 59.1,
     "unite": "%",
     "reflexion": "budget",
     "sortie_modele": "2025-11-24",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Letta Code"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/letta-code/"
    },
    {
     "harnais": "mux",
     "nom": "Xum (ex-Mux)",
     "type": "grand_public",
     "pas": null,
     "valeur": 58.43,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-11-24",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Mux"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/mux/"
    },
    {
     "harnais": "terminus-2",
     "nom": "Terminus 2",
     "type": "banc",
     "pas": null,
     "valeur": 57.75,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-11-24",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Terminus 2"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/terminus-2/"
    },
    {
     "harnais": "goose",
     "nom": "goose",
     "type": "grand_public",
     "pas": null,
     "valeur": 54.33,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-11-24",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Goose"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/goose/"
    },
    {
     "harnais": "claude-code",
     "nom": "Claude Code",
     "type": "grand_public",
     "pas": null,
     "valeur": 52.13,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-11-24",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Claude Code"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/claude-code/"
    },
    {
     "harnais": "openhands",
     "nom": "OpenHands",
     "type": "grand_public",
     "pas": null,
     "valeur": 51.91,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-11-24",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "OpenHands"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/openhands/"
    },
    {
     "harnais": "opencode",
     "nom": "OpenCode",
     "type": "grand_public",
     "pas": null,
     "valeur": 51.69,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-11-24",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "OpenCode"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/opencode/"
    }
   ],
   "retirees": []
  },
  {
   "id": "terminal_bench--gemini-3-pro",
   "modele": "gemini-3-pro",
   "modele_nom": "Gemini 3 Pro",
   "modele_url": "https://quelleia.com/modeles/gemini-3-pro/",
   "benchmark": "terminal_bench",
   "benchmark_nom": "Terminal-Bench 2.0",
   "benchmark_url": "https://quelleia.com/benchmarks/terminal_bench/",
   "nature": "harnais",
   "harnais": 7,
   "ecart": 13.47,
   "barres": [
    {
     "harnais": "ante",
     "nom": "Ante",
     "type": "autre",
     "pas": null,
     "valeur": 69.43,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-11-18",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Ante"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/ante/"
    },
    {
     "harnais": "sageagent",
     "nom": "SageAgent",
     "type": "autre",
     "pas": null,
     "valeur": 65.17,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-11-18",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "SageAgent"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/sageagent/"
    },
    {
     "harnais": "codebrain-1-5",
     "nom": "CodeBrain-1.5",
     "type": "autre",
     "pas": null,
     "valeur": 62.25,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-11-18",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "CodeBrain-1",
      "CodeBrain-1.5"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/codebrain-1-5/"
    },
    {
     "harnais": "ii-agent",
     "nom": "II-Agent",
     "type": "autre",
     "pas": null,
     "valeur": 61.8,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-11-18",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "II-Agent"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/ii-agent/"
    },
    {
     "harnais": "droid",
     "nom": "Droid (Factory)",
     "type": "grand_public",
     "pas": null,
     "valeur": 61.12,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-11-18",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Droid"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/factory-droid/"
    },
    {
     "harnais": "terminus-2",
     "nom": "Terminus 2",
     "type": "banc",
     "pas": null,
     "valeur": 56.85,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-11-18",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Terminus 2"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/terminus-2/"
    },
    {
     "harnais": "letta-code",
     "nom": "Letta Code",
     "type": "grand_public",
     "pas": null,
     "valeur": 55.96,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-11-18",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Letta Code"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/letta-code/"
    }
   ],
   "retirees": []
  },
  {
   "id": "terminal_bench--claude-sonnet-4-5",
   "modele": "claude-sonnet-4-5",
   "modele_nom": "Claude Sonnet 4.5",
   "modele_url": "https://quelleia.com/modeles/claude-sonnet-4-5/",
   "benchmark": "terminal_bench",
   "benchmark_nom": "Terminal-Bench 2.0",
   "benchmark_url": "https://quelleia.com/benchmarks/terminal_bench/",
   "nature": "harnais",
   "harnais": 7,
   "ecart": 6.46,
   "barres": [
    {
     "harnais": "camel-ai",
     "nom": "CAMEL-AI",
     "type": "recherche",
     "pas": null,
     "valeur": 46.52,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-09-29",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "CAMEL-AI"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/camel-ai/"
    },
    {
     "harnais": "goose",
     "nom": "goose",
     "type": "grand_public",
     "pas": null,
     "valeur": 43.15,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-09-29",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Goose"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/goose/"
    },
    {
     "harnais": "terminus-2",
     "nom": "Terminus 2",
     "type": "banc",
     "pas": null,
     "valeur": 42.81,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-09-29",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Terminus 2"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/terminus-2/"
    },
    {
     "harnais": "maya-v2",
     "nom": "MAYA-V2",
     "type": "autre",
     "pas": null,
     "valeur": 42.7,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-09-29",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "MAYA",
      "MAYA-V2"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/maya-v2/"
    },
    {
     "harnais": "openhands",
     "nom": "OpenHands",
     "type": "grand_public",
     "pas": null,
     "valeur": 42.58,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-09-29",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "OpenHands"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/openhands/"
    },
    {
     "harnais": "mini-swe-agent",
     "nom": "mini-SWE-agent",
     "type": "banc",
     "pas": null,
     "valeur": 42.53,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-09-29",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "mini-SWE-agent"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/mini-swe-agent/"
    },
    {
     "harnais": "claude-code",
     "nom": "Claude Code",
     "type": "grand_public",
     "pas": null,
     "valeur": 40.06,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-09-29",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Claude Code"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/claude-code/"
    }
   ],
   "retirees": []
  },
  {
   "id": "terminal_bench--claude-haiku-4-5",
   "modele": "claude-haiku-4-5",
   "modele_nom": "Claude Haiku 4.5",
   "modele_url": "https://quelleia.com/modeles/claude-haiku-4-5/",
   "benchmark": "terminal_bench",
   "benchmark_nom": "Terminal-Bench 2.0",
   "benchmark_url": "https://quelleia.com/benchmarks/terminal_bench/",
   "nature": "harnais",
   "harnais": 5,
   "ecart": 21.57,
   "barres": [
    {
     "harnais": "goose",
     "nom": "goose",
     "type": "grand_public",
     "pas": null,
     "valeur": 35.51,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-10-15",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Goose"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/goose/"
    },
    {
     "harnais": "mini-swe-agent",
     "nom": "mini-SWE-agent",
     "type": "banc",
     "pas": null,
     "valeur": 29.83,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-10-15",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "mini-SWE-agent"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/mini-swe-agent/"
    },
    {
     "harnais": "terminus-2",
     "nom": "Terminus 2",
     "type": "banc",
     "pas": null,
     "valeur": 28.31,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-10-15",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Terminus 2"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/terminus-2/"
    },
    {
     "harnais": "claude-code",
     "nom": "Claude Code",
     "type": "grand_public",
     "pas": null,
     "valeur": 27.53,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-10-15",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Claude Code"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/claude-code/"
    },
    {
     "harnais": "openhands",
     "nom": "OpenHands",
     "type": "grand_public",
     "pas": null,
     "valeur": 13.93,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-10-15",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "OpenHands"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/openhands/"
    }
   ],
   "retirees": []
  },
  {
   "id": "terminal_bench--gpt-5-nano",
   "modele": "gpt-5-nano",
   "modele_nom": "GPT-5 nano",
   "modele_url": "https://quelleia.com/modeles/gpt-5-nano/",
   "benchmark": "terminal_bench",
   "benchmark_nom": "Terminal-Bench 2.0",
   "benchmark_url": "https://quelleia.com/benchmarks/terminal_bench/",
   "nature": "harnais",
   "harnais": 5,
   "ecart": 14.83,
   "barres": [
    {
     "harnais": "spoox-o-m",
     "nom": "spoox-o-m",
     "type": "recherche",
     "pas": null,
     "valeur": 21.8,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-08-07",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "spoox-o-m"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/spoox-o-m/"
    },
    {
     "harnais": "codex-cli",
     "nom": "Codex CLI",
     "type": "grand_public",
     "pas": null,
     "valeur": 11.46,
     "unite": "%",
     "reflexion": "moyenne",
     "sortie_modele": "2025-08-07",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Codex CLI"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/codex/"
    },
    {
     "harnais": "openhands",
     "nom": "OpenHands",
     "type": "grand_public",
     "pas": null,
     "valeur": 9.89,
     "unite": "%",
     "reflexion": "moyenne",
     "sortie_modele": "2025-08-07",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "OpenHands"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/openhands/"
    },
    {
     "harnais": "terminus-2",
     "nom": "Terminus 2",
     "type": "banc",
     "pas": null,
     "valeur": 7.87,
     "unite": "%",
     "reflexion": "moyenne",
     "sortie_modele": "2025-08-07",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Terminus 2"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/terminus-2/"
    },
    {
     "harnais": "mini-swe-agent",
     "nom": "mini-SWE-agent",
     "type": "banc",
     "pas": null,
     "valeur": 6.97,
     "unite": "%",
     "reflexion": "moyenne",
     "sortie_modele": "2025-08-07",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "mini-SWE-agent"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/mini-swe-agent/"
    }
   ],
   "retirees": []
  },
  {
   "id": "terminal_bench--gpt-5-mini",
   "modele": "gpt-5-mini",
   "modele_nom": "GPT-5 mini",
   "modele_url": "https://quelleia.com/modeles/gpt-5-mini/",
   "benchmark": "terminal_bench",
   "benchmark_nom": "Terminal-Bench 2.0",
   "benchmark_url": "https://quelleia.com/benchmarks/terminal_bench/",
   "nature": "harnais",
   "harnais": 5,
   "ecart": 12.58,
   "barres": [
    {
     "harnais": "spoox-o-m",
     "nom": "spoox-o-m",
     "type": "recherche",
     "pas": null,
     "valeur": 34.83,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-08-07",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "spoox-m",
      "spoox-o-m"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/spoox-o-m/"
    },
    {
     "harnais": "codex-cli",
     "nom": "Codex CLI",
     "type": "grand_public",
     "pas": null,
     "valeur": 31.85,
     "unite": "%",
     "reflexion": "moyenne",
     "sortie_modele": "2025-08-07",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Codex CLI"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/codex/"
    },
    {
     "harnais": "openhands",
     "nom": "OpenHands",
     "type": "grand_public",
     "pas": null,
     "valeur": 29.21,
     "unite": "%",
     "reflexion": "moyenne",
     "sortie_modele": "2025-08-07",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "OpenHands"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/openhands/"
    },
    {
     "harnais": "terminus-2",
     "nom": "Terminus 2",
     "type": "banc",
     "pas": null,
     "valeur": 24.04,
     "unite": "%",
     "reflexion": "moyenne",
     "sortie_modele": "2025-08-07",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Terminus 2"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/terminus-2/"
    },
    {
     "harnais": "mini-swe-agent",
     "nom": "mini-SWE-agent",
     "type": "banc",
     "pas": null,
     "valeur": 22.25,
     "unite": "%",
     "reflexion": "moyenne",
     "sortie_modele": "2025-08-07",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "mini-SWE-agent"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/mini-swe-agent/"
    }
   ],
   "retirees": []
  },
  {
   "id": "terminal_bench--gpt-5-5",
   "modele": "gpt-5-5",
   "modele_nom": "GPT-5.5",
   "modele_url": "https://quelleia.com/modeles/gpt-5-5/",
   "benchmark": "terminal_bench",
   "benchmark_nom": "Terminal-Bench 2.0",
   "benchmark_url": "https://quelleia.com/benchmarks/terminal_bench/",
   "nature": "harnais",
   "harnais": 4,
   "ecart": 18.65,
   "barres": [
    {
     "harnais": "nexau-ahe",
     "nom": "NexAU-AHE",
     "type": "autre",
     "pas": null,
     "valeur": 84.72,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2026-04-23",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "NexAU-AHE"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/nexau-ahe/"
    },
    {
     "harnais": "capy",
     "nom": "Capy",
     "type": "autre",
     "pas": null,
     "valeur": 83.15,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2026-04-23",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Capy"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/capy/"
    },
    {
     "harnais": "codex-cli",
     "nom": "Codex CLI",
     "type": "grand_public",
     "pas": null,
     "valeur": 82.25,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2026-04-23",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Codex CLI"
     ],
     "autres_soumissions": [
      82
     ],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/codex/"
    },
    {
     "harnais": "clnkr",
     "nom": "clnkr",
     "type": "autre",
     "pas": null,
     "valeur": 66.07,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2026-04-23",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "clnkr"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/clnkr/"
    }
   ],
   "retirees": []
  },
  {
   "id": "terminal_bench--gemini-2-5-pro-jun-2025",
   "modele": "gemini-2-5-pro-jun-2025",
   "modele_nom": "Gemini 2.5 Pro (Jun 2025)",
   "modele_url": "https://quelleia.com/modeles/gemini-2-5-pro-jun-2025/",
   "benchmark": "terminal_bench",
   "benchmark_nom": "Terminal-Bench 2.0",
   "benchmark_url": "https://quelleia.com/benchmarks/terminal_bench/",
   "nature": "harnais",
   "harnais": 4,
   "ecart": 16.24,
   "barres": [
    {
     "harnais": "terminus-2",
     "nom": "Terminus 2",
     "type": "banc",
     "pas": null,
     "valeur": 32.64,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-06-17",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Terminus 2"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/terminus-2/"
    },
    {
     "harnais": "mini-swe-agent",
     "nom": "mini-SWE-agent",
     "type": "banc",
     "pas": null,
     "valeur": 26.07,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-06-17",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "mini-SWE-agent"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/mini-swe-agent/"
    },
    {
     "harnais": "gemini-cli",
     "nom": "Gemini CLI",
     "type": "grand_public",
     "pas": null,
     "valeur": 19.55,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-06-17",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Gemini CLI"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/gemini-cli/"
    },
    {
     "harnais": "openhands",
     "nom": "OpenHands",
     "type": "grand_public",
     "pas": null,
     "valeur": 16.4,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-06-17",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "OpenHands"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/openhands/"
    }
   ],
   "retirees": []
  },
  {
   "id": "terminal_bench--gpt-5",
   "modele": "gpt-5",
   "modele_nom": "GPT-5",
   "modele_url": "https://quelleia.com/modeles/gpt-5/",
   "benchmark": "terminal_bench",
   "benchmark_nom": "Terminal-Bench 2.0",
   "benchmark_url": "https://quelleia.com/benchmarks/terminal_bench/",
   "nature": "harnais",
   "harnais": 4,
   "ecart": 15.67,
   "barres": [
    {
     "harnais": "codex-cli",
     "nom": "Codex CLI",
     "type": "grand_public",
     "pas": null,
     "valeur": 49.61,
     "unite": "%",
     "reflexion": "moyenne",
     "sortie_modele": "2025-08-07",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Codex CLI"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/codex/"
    },
    {
     "harnais": "openhands",
     "nom": "OpenHands",
     "type": "grand_public",
     "pas": null,
     "valeur": 43.82,
     "unite": "%",
     "reflexion": "moyenne",
     "sortie_modele": "2025-08-07",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "OpenHands"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/openhands/"
    },
    {
     "harnais": "terminus-2",
     "nom": "Terminus 2",
     "type": "banc",
     "pas": null,
     "valeur": 35.17,
     "unite": "%",
     "reflexion": "moyenne",
     "sortie_modele": "2025-08-07",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Terminus 2"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/terminus-2/"
    },
    {
     "harnais": "mini-swe-agent",
     "nom": "mini-SWE-agent",
     "type": "banc",
     "pas": null,
     "valeur": 33.93,
     "unite": "%",
     "reflexion": "moyenne",
     "sortie_modele": "2025-08-07",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "mini-SWE-agent"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/mini-swe-agent/"
    }
   ],
   "retirees": []
  },
  {
   "id": "terminal_bench--gpt-5-2",
   "modele": "gpt-5-2",
   "modele_nom": "GPT-5.2",
   "modele_url": "https://quelleia.com/modeles/gpt-5-2/",
   "benchmark": "terminal_bench",
   "benchmark_nom": "Terminal-Bench 2.0",
   "benchmark_url": "https://quelleia.com/benchmarks/terminal_bench/",
   "nature": "harnais",
   "harnais": 4,
   "ecart": 10.9,
   "barres": [
    {
     "harnais": "droid",
     "nom": "Droid (Factory)",
     "type": "grand_public",
     "pas": null,
     "valeur": 64.94,
     "unite": "%",
     "reflexion": "moyenne",
     "sortie_modele": "2025-12-11",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Droid"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/factory-droid/"
    },
    {
     "harnais": "codex-cli",
     "nom": "Codex CLI",
     "type": "grand_public",
     "pas": null,
     "valeur": 62.92,
     "unite": "%",
     "reflexion": "moyenne",
     "sortie_modele": "2025-12-11",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Codex CLI"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/codex/"
    },
    {
     "harnais": "mux",
     "nom": "Xum (ex-Mux)",
     "type": "grand_public",
     "pas": null,
     "valeur": 60.67,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-12-11",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Mux"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/mux/"
    },
    {
     "harnais": "terminus-2",
     "nom": "Terminus 2",
     "type": "banc",
     "pas": null,
     "valeur": 54.04,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-12-11",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Terminus 2"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/terminus-2/"
    }
   ],
   "retirees": []
  },
  {
   "id": "terminal_bench--claude-opus-4-1",
   "modele": "claude-opus-4-1",
   "modele_nom": "Claude Opus 4.1",
   "modele_url": "https://quelleia.com/modeles/claude-opus-4-1/",
   "benchmark": "terminal_bench",
   "benchmark_nom": "Terminal-Bench 2.0",
   "benchmark_url": "https://quelleia.com/benchmarks/terminal_bench/",
   "nature": "harnais",
   "harnais": 4,
   "ecart": 3.15,
   "barres": [
    {
     "harnais": "terminus-2",
     "nom": "Terminus 2",
     "type": "banc",
     "pas": null,
     "valeur": 37.98,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-08-05",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Terminus 2"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/terminus-2/"
    },
    {
     "harnais": "openhands",
     "nom": "OpenHands",
     "type": "grand_public",
     "pas": null,
     "valeur": 36.85,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-08-05",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "OpenHands"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/openhands/"
    },
    {
     "harnais": "mini-swe-agent",
     "nom": "mini-SWE-agent",
     "type": "banc",
     "pas": null,
     "valeur": 35.06,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-08-05",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "mini-SWE-agent"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/mini-swe-agent/"
    },
    {
     "harnais": "claude-code",
     "nom": "Claude Code",
     "type": "grand_public",
     "pas": null,
     "valeur": 34.83,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-08-05",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Claude Code"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/claude-code/"
    }
   ],
   "retirees": []
  },
  {
   "id": "terminal_bench--gemini-2-5-flash-sep-2025",
   "modele": "gemini-2-5-flash-sep-2025",
   "modele_nom": "Gemini 2.5 Flash (Sep 2025)",
   "modele_url": "https://quelleia.com/modeles/gemini-2-5-flash-sep-2025/",
   "benchmark": "terminal_bench",
   "benchmark_nom": "Terminal-Bench 2.0",
   "benchmark_url": "https://quelleia.com/benchmarks/terminal_bench/",
   "nature": "harnais",
   "harnais": 4,
   "ecart": 1.7,
   "barres": [
    {
     "harnais": "mini-swe-agent",
     "nom": "mini-SWE-agent",
     "type": "banc",
     "pas": null,
     "valeur": 17.1,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-11-03",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "mini-SWE-agent"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/mini-swe-agent/"
    },
    {
     "harnais": "terminus-2",
     "nom": "Terminus 2",
     "type": "banc",
     "pas": null,
     "valeur": 16.9,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-10-31",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Terminus 2"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/terminus-2/"
    },
    {
     "harnais": "openhands",
     "nom": "OpenHands",
     "type": "grand_public",
     "pas": null,
     "valeur": 16.4,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-11-02",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "OpenHands"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/openhands/"
    },
    {
     "harnais": "gemini-cli",
     "nom": "Gemini CLI",
     "type": "grand_public",
     "pas": null,
     "valeur": 15.4,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-11-04",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Gemini CLI"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/gemini-cli/"
    }
   ],
   "retirees": []
  },
  {
   "id": "terminal_bench--gemini-2-5-flash-jun-2025",
   "modele": "gemini-2-5-flash-jun-2025",
   "modele_nom": "Gemini 2.5 Flash (Jun 2025)",
   "modele_url": "https://quelleia.com/modeles/gemini-2-5-flash-jun-2025/",
   "benchmark": "terminal_bench",
   "benchmark_nom": "Terminal-Bench 2.0",
   "benchmark_url": "https://quelleia.com/benchmarks/terminal_bench/",
   "nature": "harnais",
   "harnais": 4,
   "ecart": 1.63,
   "barres": [
    {
     "harnais": "mini-swe-agent",
     "nom": "mini-SWE-agent",
     "type": "banc",
     "pas": null,
     "valeur": 17.08,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-06-17",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "mini-SWE-agent"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/mini-swe-agent/"
    },
    {
     "harnais": "terminus-2",
     "nom": "Terminus 2",
     "type": "banc",
     "pas": null,
     "valeur": 16.85,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-06-17",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Terminus 2"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/terminus-2/"
    },
    {
     "harnais": "openhands",
     "nom": "OpenHands",
     "type": "grand_public",
     "pas": null,
     "valeur": 16.4,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-06-17",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "OpenHands"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/openhands/"
    },
    {
     "harnais": "gemini-cli",
     "nom": "Gemini CLI",
     "type": "grand_public",
     "pas": null,
     "valeur": 15.45,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-06-17",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Gemini CLI"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/gemini-cli/"
    }
   ],
   "retirees": []
  },
  {
   "id": "terminal_bench--gpt-5-1-codex",
   "modele": "gpt-5-1-codex",
   "modele_nom": "GPT-5.1-Codex",
   "modele_url": "https://quelleia.com/modeles/gpt-5-1-codex/",
   "benchmark": "terminal_bench",
   "benchmark_nom": "Terminal-Bench 2.0",
   "benchmark_url": "https://quelleia.com/benchmarks/terminal_bench/",
   "nature": "harnais",
   "harnais": 3,
   "ecart": 20.9,
   "barres": [
    {
     "harnais": "crux",
     "nom": "Crux",
     "type": "autre",
     "pas": null,
     "valeur": 57.75,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-11-12",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Crux"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/crux/"
    },
    {
     "harnais": "letta-code",
     "nom": "Letta Code",
     "type": "grand_public",
     "pas": null,
     "valeur": 53.48,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-11-12",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Letta Code"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/letta-code/"
    },
    {
     "harnais": "terminus-2",
     "nom": "Terminus 2",
     "type": "banc",
     "pas": null,
     "valeur": 36.85,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-11-12",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Terminus 2"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/terminus-2/"
    }
   ],
   "retirees": [
    {
     "harnais": "codex-cli",
     "nom": "Codex CLI",
     "valeur": 57.8,
     "unite": "%",
     "sortie_modele": "2025-11-16",
     "constat": "Lignes présentes au classement du 26 décembre 2025 au 20 février 2026, absentes à partir du 10 mars 2026.",
     "sources": [
      "https://web.archive.org/web/20260220180020/https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "https://web.archive.org/web/20260310094521/https://www.tbench.ai/leaderboard/terminal-bench/2.0"
     ]
    }
   ]
  },
  {
   "id": "terminal_bench--gemini-3-1-pro",
   "modele": "gemini-3-1-pro",
   "modele_nom": "Gemini 3.1 Pro",
   "modele_url": "https://quelleia.com/modeles/gemini-3-1-pro/",
   "benchmark": "terminal_bench",
   "benchmark_nom": "Terminal-Bench 2.0",
   "benchmark_url": "https://quelleia.com/benchmarks/terminal_bench/",
   "nature": "harnais",
   "harnais": 3,
   "ecart": 20.86,
   "barres": [
    {
     "harnais": "tongagents",
     "nom": "TongAgents",
     "type": "recherche",
     "pas": null,
     "valeur": 80.22,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2026-02-19",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "TongAgents"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/tongagents/"
    },
    {
     "harnais": "terminus-kira",
     "nom": "Terminus-KIRA",
     "type": "autre",
     "pas": null,
     "valeur": 74.83,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2026-02-19",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Terminus-KIRA"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/terminus-kira/"
    },
    {
     "harnais": "gemini-cli",
     "nom": "Gemini CLI",
     "type": "grand_public",
     "pas": null,
     "valeur": 59.36,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2026-02-19",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Gemini CLI"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/gemini-cli/"
    }
   ],
   "retirees": [
    {
     "harnais": "forgecode",
     "nom": "ForgeCode",
     "valeur": 78.4,
     "unite": "%",
     "sortie_modele": "2026-03-02",
     "constat": "Les trois lignes de ForgeCode figurent au classement Terminal-Bench 2.0 jusqu'au 13 mai 2026 et n'y sont plus à partir du 19 mai 2026. Le 19 avril 2026, l'équipe du classement écrivait avoir remis à zéro des essais où l'agent de ForgeCode récupérait la solution sur Internet.",
     "sources": [
      "https://web.archive.org/web/20260513024259/https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "https://web.archive.org/web/20260519111531/https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "https://www.tbench.ai/news/leaderboard-integrity-update"
     ]
    }
   ]
  },
  {
   "id": "terminal_bench--gemini-3-flash",
   "modele": "gemini-3-flash",
   "modele_nom": "Gemini 3 Flash",
   "modele_url": "https://quelleia.com/modeles/gemini-3-flash/",
   "benchmark": "terminal_bench",
   "benchmark_nom": "Terminal-Bench 2.0",
   "benchmark_url": "https://quelleia.com/benchmarks/terminal_bench/",
   "nature": "harnais",
   "harnais": 3,
   "ecart": 16.85,
   "barres": [
    {
     "harnais": "junie-cli",
     "nom": "Junie CLI",
     "type": "grand_public",
     "pas": null,
     "valeur": 64.27,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-12-17",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Junie CLI"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/junie/"
    },
    {
     "harnais": "terminus-2",
     "nom": "Terminus 2",
     "type": "banc",
     "pas": null,
     "valeur": 51.69,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-12-17",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Terminus 2"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/terminus-2/"
    },
    {
     "harnais": "gemini-cli",
     "nom": "Gemini CLI",
     "type": "grand_public",
     "pas": null,
     "valeur": 47.42,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-12-17",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Gemini CLI"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/gemini-cli/"
    }
   ],
   "retirees": [
    {
     "harnais": "gemini-cli",
     "nom": "Gemini CLI",
     "valeur": 51,
     "unite": "%",
     "sortie_modele": "2025-12-23",
     "constat": "Ligne présente jusqu'au 10 mars 2026, remplacée par la mesure à 47,4 % (ligne du 6 mars 2026), seule restée au classement.",
     "sources": [
      "https://web.archive.org/web/20260310094521/https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "https://web.archive.org/web/20260315050912/https://www.tbench.ai/leaderboard/terminal-bench/2.0"
     ]
    }
   ]
  },
  {
   "id": "terminal_bench--grok-4",
   "modele": "grok-4",
   "modele_nom": "Grok 4",
   "modele_url": "https://quelleia.com/modeles/grok-4/",
   "benchmark": "terminal_bench",
   "benchmark_nom": "Terminal-Bench 2.0",
   "benchmark_url": "https://quelleia.com/benchmarks/terminal_bench/",
   "nature": "harnais",
   "harnais": 3,
   "ecart": 4.04,
   "barres": [
    {
     "harnais": "openhands",
     "nom": "OpenHands",
     "type": "grand_public",
     "pas": null,
     "valeur": 27.19,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-07-09",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "OpenHands"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/openhands/"
    },
    {
     "harnais": "mini-swe-agent",
     "nom": "mini-SWE-agent",
     "type": "banc",
     "pas": null,
     "valeur": 25.39,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-07-09",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "mini-SWE-agent"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/mini-swe-agent/"
    },
    {
     "harnais": "terminus-2",
     "nom": "Terminus 2",
     "type": "banc",
     "pas": null,
     "valeur": 23.15,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-07-09",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Terminus 2"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/terminus-2/"
    }
   ],
   "retirees": []
  },
  {
   "id": "terminal_bench--qwen3-coder-480b-a35b",
   "modele": "qwen3-coder-480b-a35b",
   "modele_nom": "Qwen3-Coder-480B-A35B",
   "modele_url": "https://quelleia.com/modeles/qwen3-coder-480b-a35b/",
   "benchmark": "terminal_bench",
   "benchmark_nom": "Terminal-Bench 2.0",
   "benchmark_url": "https://quelleia.com/benchmarks/terminal_bench/",
   "nature": "harnais",
   "harnais": 3,
   "ecart": 3.3,
   "barres": [
    {
     "harnais": "dakou-agent",
     "nom": "Dakou Agent",
     "type": "autre",
     "pas": null,
     "valeur": 27.19,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-07-22",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Dakou Agent"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/dakou-agent/"
    },
    {
     "harnais": "openhands",
     "nom": "OpenHands",
     "type": "grand_public",
     "pas": null,
     "valeur": 25.39,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-07-22",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "OpenHands"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/openhands/"
    },
    {
     "harnais": "terminus-2",
     "nom": "Terminus 2",
     "type": "banc",
     "pas": null,
     "valeur": 23.9,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-07-22",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Terminus 2"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/terminus-2/"
    }
   ],
   "retirees": []
  },
  {
   "id": "terminal_bench--gpt-5-codex",
   "modele": "gpt-5-codex",
   "modele_nom": "GPT-5-Codex",
   "modele_url": "https://quelleia.com/modeles/gpt-5-codex/",
   "benchmark": "terminal_bench",
   "benchmark_nom": "Terminal-Bench 2.0",
   "benchmark_url": "https://quelleia.com/benchmarks/terminal_bench/",
   "nature": "harnais",
   "harnais": 3,
   "ecart": 2.98,
   "barres": [
    {
     "harnais": "codex-cli",
     "nom": "Codex CLI",
     "type": "grand_public",
     "pas": null,
     "valeur": 44.33,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-09-15",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Codex CLI"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/codex/"
    },
    {
     "harnais": "terminus-2",
     "nom": "Terminus 2",
     "type": "banc",
     "pas": null,
     "valeur": 43.43,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-09-15",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Terminus 2"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/terminus-2/"
    },
    {
     "harnais": "mini-swe-agent",
     "nom": "mini-SWE-agent",
     "type": "banc",
     "pas": null,
     "valeur": 41.35,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-09-15",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "mini-SWE-agent"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/mini-swe-agent/"
    }
   ],
   "retirees": []
  },
  {
   "id": "terminal_bench--gpt-5-1-codex-mini",
   "modele": "gpt-5-1-codex-mini",
   "modele_nom": "GPT-5.1-Codex-mini",
   "modele_url": "https://quelleia.com/modeles/gpt-5-1-codex-mini/",
   "benchmark": "terminal_bench",
   "benchmark_nom": "Terminal-Bench 2.0",
   "benchmark_url": "https://quelleia.com/benchmarks/terminal_bench/",
   "nature": "harnais",
   "harnais": 2,
   "ecart": 18.43,
   "barres": [
    {
     "harnais": "hookele",
     "nom": "hookele",
     "type": "autre",
     "pas": null,
     "valeur": 61.57,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-11-12",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "hookele"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/hookele/"
    },
    {
     "harnais": "crux",
     "nom": "Crux",
     "type": "autre",
     "pas": null,
     "valeur": 43.15,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-11-12",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Crux"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/crux/"
    }
   ],
   "retirees": [
    {
     "harnais": "codex-cli",
     "nom": "Codex CLI",
     "valeur": 43.1,
     "unite": "%",
     "sortie_modele": "2025-11-17",
     "constat": "Lignes présentes au classement du 26 décembre 2025 au 20 février 2026, absentes à partir du 10 mars 2026.",
     "sources": [
      "https://web.archive.org/web/20260220180020/https://www.tbench.ai/leaderboard/terminal-bench/2.0",
      "https://web.archive.org/web/20260310094521/https://www.tbench.ai/leaderboard/terminal-bench/2.0"
     ]
    }
   ]
  },
  {
   "id": "terminal_bench--grok-code-fast-1",
   "modele": "grok-code-fast-1",
   "modele_nom": "grok-code-fast-1",
   "modele_url": "https://quelleia.com/modeles/grok-code-fast-1/",
   "benchmark": "terminal_bench",
   "benchmark_nom": "Terminal-Bench 2.0",
   "benchmark_url": "https://quelleia.com/benchmarks/terminal_bench/",
   "nature": "harnais",
   "harnais": 2,
   "ecart": 11.69,
   "barres": [
    {
     "harnais": "mini-swe-agent",
     "nom": "mini-SWE-agent",
     "type": "banc",
     "pas": null,
     "valeur": 25.84,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-08-28",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "mini-SWE-agent"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/mini-swe-agent/"
    },
    {
     "harnais": "terminus-2",
     "nom": "Terminus 2",
     "type": "banc",
     "pas": null,
     "valeur": 14.16,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-08-28",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Terminus 2"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/terminus-2/"
    }
   ],
   "retirees": []
  },
  {
   "id": "terminal_bench--minimax-m2-1",
   "modele": "minimax-m2-1",
   "modele_nom": "MiniMax-M2.1",
   "modele_url": "https://quelleia.com/modeles/minimax-m2-1/",
   "benchmark": "terminal_bench",
   "benchmark_nom": "Terminal-Bench 2.0",
   "benchmark_url": "https://quelleia.com/benchmarks/terminal_bench/",
   "nature": "harnais",
   "harnais": 2,
   "ecart": 7.42,
   "barres": [
    {
     "harnais": "crux",
     "nom": "Crux",
     "type": "autre",
     "pas": null,
     "valeur": 36.63,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-12-23",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Crux"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/crux/"
    },
    {
     "harnais": "terminus-2",
     "nom": "Terminus 2",
     "type": "banc",
     "pas": null,
     "valeur": 29.21,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-12-23",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Terminus 2"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/terminus-2/"
    }
   ],
   "retirees": []
  },
  {
   "id": "terminal_bench--gpt-oss-120b",
   "modele": "gpt-oss-120b",
   "modele_nom": "gpt-oss-120b",
   "modele_url": "https://quelleia.com/modeles/gpt-oss-120b/",
   "benchmark": "terminal_bench",
   "benchmark_nom": "Terminal-Bench 2.0",
   "benchmark_url": "https://quelleia.com/benchmarks/terminal_bench/",
   "nature": "harnais",
   "harnais": 2,
   "ecart": 4.53,
   "barres": [
    {
     "harnais": "terminus-2",
     "nom": "Terminus 2",
     "type": "banc",
     "pas": null,
     "valeur": 18.69,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-08-05",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Terminus 2"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/terminus-2/"
    },
    {
     "harnais": "mini-swe-agent",
     "nom": "mini-SWE-agent",
     "type": "banc",
     "pas": null,
     "valeur": 14.16,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-08-05",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "mini-SWE-agent"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/mini-swe-agent/"
    }
   ],
   "retirees": []
  },
  {
   "id": "terminal_bench--minimax-m2-7",
   "modele": "minimax-m2-7",
   "modele_nom": "MiniMax-M2.7",
   "modele_url": "https://quelleia.com/modeles/minimax-m2-7/",
   "benchmark": "terminal_bench",
   "benchmark_nom": "Terminal-Bench 2.0",
   "benchmark_url": "https://quelleia.com/benchmarks/terminal_bench/",
   "nature": "harnais",
   "harnais": 2,
   "ecart": 2.13,
   "barres": [
    {
     "harnais": "indusagi",
     "nom": "IndusAGI Coding Agent",
     "type": "autre",
     "pas": null,
     "valeur": 45.06,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2026-03-18",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "IndusAGI Coding Agent"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/indusagi/"
    },
    {
     "harnais": "harness-agent",
     "nom": "Harness Agent",
     "type": "autre",
     "pas": null,
     "valeur": 42.92,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2026-03-18",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Harness Agent"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/harness-agent/"
    }
   ],
   "retirees": []
  },
  {
   "id": "terminal_bench--kimi-k2-jul-2025",
   "modele": "kimi-k2-jul-2025",
   "modele_nom": "Kimi K2 (Jul 2025)",
   "modele_url": "https://quelleia.com/modeles/kimi-k2-jul-2025/",
   "benchmark": "terminal_bench",
   "benchmark_nom": "Terminal-Bench 2.0",
   "benchmark_url": "https://quelleia.com/benchmarks/terminal_bench/",
   "nature": "harnais",
   "harnais": 2,
   "ecart": 1.05,
   "barres": [
    {
     "harnais": "terminus-2",
     "nom": "Terminus 2",
     "type": "banc",
     "pas": null,
     "valeur": 27.79,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-07-11",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Terminus 2"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/terminus-2/"
    },
    {
     "harnais": "openhands",
     "nom": "OpenHands",
     "type": "grand_public",
     "pas": null,
     "valeur": 26.74,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-07-11",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "OpenHands"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/openhands/"
    }
   ],
   "retirees": []
  },
  {
   "id": "terminal_bench--minimax-m2-5",
   "modele": "minimax-m2-5",
   "modele_nom": "MiniMax-M2.5",
   "modele_url": "https://quelleia.com/modeles/minimax-m2-5/",
   "benchmark": "terminal_bench",
   "benchmark_nom": "Terminal-Bench 2.0",
   "benchmark_url": "https://quelleia.com/benchmarks/terminal_bench/",
   "nature": "harnais",
   "harnais": 2,
   "ecart": 0.45,
   "barres": [
    {
     "harnais": "cchuter",
     "nom": "cchuter",
     "type": "autre",
     "pas": null,
     "valeur": 42.7,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2026-02-12",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "cchuter"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/cchuter/"
    },
    {
     "harnais": "terminus-2",
     "nom": "Terminus 2",
     "type": "banc",
     "pas": null,
     "valeur": 42.25,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2026-02-12",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Terminus 2"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/terminus-2/"
    }
   ],
   "retirees": []
  },
  {
   "id": "terminal_bench--gpt-oss-20b",
   "modele": "gpt-oss-20b",
   "modele_nom": "gpt-oss-20b",
   "modele_url": "https://quelleia.com/modeles/gpt-oss-20b/",
   "benchmark": "terminal_bench",
   "benchmark_nom": "Terminal-Bench 2.0",
   "benchmark_url": "https://quelleia.com/benchmarks/terminal_bench/",
   "nature": "harnais",
   "harnais": 2,
   "ecart": 0.3,
   "barres": [
    {
     "harnais": "mini-swe-agent",
     "nom": "mini-SWE-agent",
     "type": "banc",
     "pas": null,
     "valeur": 3.37,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-08-05",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "mini-SWE-agent"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/mini-swe-agent/"
    },
    {
     "harnais": "terminus-2",
     "nom": "Terminus 2",
     "type": "banc",
     "pas": null,
     "valeur": 3.07,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-08-05",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Terminus 2"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 1,
     "fiche": "https://quelleia.com/outils/terminus-2/"
    }
   ],
   "retirees": []
  },
  {
   "id": "terminal_bench--glm-4-7",
   "modele": "glm-4-7",
   "modele_nom": "GLM-4.7",
   "modele_url": "https://quelleia.com/modeles/glm-4-7/",
   "benchmark": "terminal_bench",
   "benchmark_nom": "Terminal-Bench 2.0",
   "benchmark_url": "https://quelleia.com/benchmarks/terminal_bench/",
   "nature": "harnais",
   "harnais": 2,
   "ecart": 0.13,
   "barres": [
    {
     "harnais": "terminus-2",
     "nom": "Terminus 2",
     "type": "banc",
     "pas": null,
     "valeur": 33.39,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-12-22",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Terminus 2"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/terminus-2/"
    },
    {
     "harnais": "crux",
     "nom": "Crux",
     "type": "autre",
     "pas": null,
     "valeur": 33.26,
     "unite": "%",
     "reflexion": "inconnue",
     "sortie_modele": "2025-12-22",
     "source": "https://www.tbench.ai/leaderboard/terminal-bench/2.0",
     "ecritures": [
      "Crux"
     ],
     "autres_soumissions": [],
     "doublons_ecartes": 0,
     "fiche": "https://quelleia.com/outils/crux/"
    }
   ],
   "retirees": []
  }
 ],
 "retirees": [
  {
   "modele": "gpt-5-4",
   "benchmark": "terminal_bench",
   "harnais": "forgecode",
   "nom": "ForgeCode",
   "valeur": 81.8,
   "unite": "%",
   "sortie_modele": "2026-03-12",
   "constat": "Les trois lignes de ForgeCode figurent au classement Terminal-Bench 2.0 jusqu'au 13 mai 2026 et n'y sont plus à partir du 19 mai 2026. Le 19 avril 2026, l'équipe du classement écrivait avoir remis à zéro des essais où l'agent de ForgeCode récupérait la solution sur Internet.",
   "sources": [
    "https://web.archive.org/web/20260513024259/https://www.tbench.ai/leaderboard/terminal-bench/2.0",
    "https://web.archive.org/web/20260519111531/https://www.tbench.ai/leaderboard/terminal-bench/2.0",
    "https://www.tbench.ai/news/leaderboard-integrity-update"
   ]
  },
  {
   "modele": "claude-opus-4-6",
   "benchmark": "terminal_bench",
   "harnais": "forgecode",
   "nom": "ForgeCode",
   "valeur": 79.8,
   "unite": "%",
   "sortie_modele": "2026-03-12",
   "constat": "Les trois lignes de ForgeCode figurent au classement Terminal-Bench 2.0 jusqu'au 13 mai 2026 et n'y sont plus à partir du 19 mai 2026. Le 19 avril 2026, l'équipe du classement écrivait avoir remis à zéro des essais où l'agent de ForgeCode récupérait la solution sur Internet.",
   "sources": [
    "https://web.archive.org/web/20260513024259/https://www.tbench.ai/leaderboard/terminal-bench/2.0",
    "https://web.archive.org/web/20260519111531/https://www.tbench.ai/leaderboard/terminal-bench/2.0",
    "https://www.tbench.ai/news/leaderboard-integrity-update"
   ]
  },
  {
   "modele": "gemini-3-1-pro",
   "benchmark": "terminal_bench",
   "harnais": "forgecode",
   "nom": "ForgeCode",
   "valeur": 78.4,
   "unite": "%",
   "sortie_modele": "2026-03-02",
   "constat": "Les trois lignes de ForgeCode figurent au classement Terminal-Bench 2.0 jusqu'au 13 mai 2026 et n'y sont plus à partir du 19 mai 2026. Le 19 avril 2026, l'équipe du classement écrivait avoir remis à zéro des essais où l'agent de ForgeCode récupérait la solution sur Internet.",
   "sources": [
    "https://web.archive.org/web/20260513024259/https://www.tbench.ai/leaderboard/terminal-bench/2.0",
    "https://web.archive.org/web/20260519111531/https://www.tbench.ai/leaderboard/terminal-bench/2.0",
    "https://www.tbench.ai/news/leaderboard-integrity-update"
   ]
  },
  {
   "modele": "gemini-3-flash",
   "benchmark": "terminal_bench",
   "harnais": "gemini-cli",
   "nom": "Gemini CLI",
   "valeur": 51,
   "unite": "%",
   "sortie_modele": "2025-12-23",
   "constat": "Ligne présente jusqu'au 10 mars 2026, remplacée par la mesure à 47,4 % (ligne du 6 mars 2026), seule restée au classement.",
   "sources": [
    "https://web.archive.org/web/20260310094521/https://www.tbench.ai/leaderboard/terminal-bench/2.0",
    "https://web.archive.org/web/20260315050912/https://www.tbench.ai/leaderboard/terminal-bench/2.0"
   ]
  },
  {
   "modele": "gpt-5-3-codex",
   "benchmark": "terminal_bench",
   "harnais": "codebrain-1",
   "nom": "CodeBrain-1",
   "valeur": 70.3,
   "unite": "%",
   "sortie_modele": "2026-02-10",
   "constat": "Ligne présente jusqu'au 13 mai 2026 ; le 19 mai 2026, le classement montre à sa place CodeBrain-1.5 avec GPT-5.3 Codex, à 75,8 %.",
   "sources": [
    "https://web.archive.org/web/20260513024259/https://www.tbench.ai/leaderboard/terminal-bench/2.0",
    "https://web.archive.org/web/20260519111531/https://www.tbench.ai/leaderboard/terminal-bench/2.0"
   ]
  },
  {
   "modele": "gpt-5-1-codex",
   "benchmark": "terminal_bench",
   "harnais": "codex-cli",
   "nom": "Codex CLI",
   "valeur": 57.8,
   "unite": "%",
   "sortie_modele": "2025-11-16",
   "constat": "Lignes présentes au classement du 26 décembre 2025 au 20 février 2026, absentes à partir du 10 mars 2026.",
   "sources": [
    "https://web.archive.org/web/20260220180020/https://www.tbench.ai/leaderboard/terminal-bench/2.0",
    "https://web.archive.org/web/20260310094521/https://www.tbench.ai/leaderboard/terminal-bench/2.0"
   ]
  },
  {
   "modele": "gpt-5-1-codex-mini",
   "benchmark": "terminal_bench",
   "harnais": "codex-cli",
   "nom": "Codex CLI",
   "valeur": 43.1,
   "unite": "%",
   "sortie_modele": "2025-11-17",
   "constat": "Lignes présentes au classement du 26 décembre 2025 au 20 février 2026, absentes à partir du 10 mars 2026.",
   "sources": [
    "https://web.archive.org/web/20260220180020/https://www.tbench.ai/leaderboard/terminal-bench/2.0",
    "https://web.archive.org/web/20260310094521/https://www.tbench.ai/leaderboard/terminal-bench/2.0"
   ]
  }
 ],
 "renommees": [
  {
   "de": "codebrain-1",
   "vers": "codebrain-1-5",
   "modeles": [
    "gemini-3-pro"
   ],
   "constat": "Même ligne (Gemini 3 Pro, 62,2 %, datée du 5 février 2026) affichée sous CodeBrain-1 jusqu'au 13 mai 2026, puis sous CodeBrain-1.5 à partir du 19 mai 2026.",
   "sources": [
    "https://web.archive.org/web/20260513024259/https://www.tbench.ai/leaderboard/terminal-bench/2.0",
    "https://web.archive.org/web/20260519111531/https://www.tbench.ai/leaderboard/terminal-bench/2.0"
   ]
  },
  {
   "de": "maya",
   "vers": "maya-v2",
   "modeles": [
    "claude-sonnet-4-5"
   ],
   "constat": "Même ligne (Claude Sonnet 4.5, 42,7 %, datée du 4 janvier 2026) affichée sous MAYA jusqu'au 10 mars 2026, puis sous MAYA-V2.",
   "sources": [
    "https://web.archive.org/web/20260310094521/https://www.tbench.ai/leaderboard/terminal-bench/2.0",
    "https://web.archive.org/web/20260826113700/https://www.tbench.ai/leaderboard/terminal-bench/2.0"
   ]
  },
  {
   "de": "spoox-m",
   "vers": "spoox-o-m",
   "modeles": [
    "gpt-5-mini"
   ],
   "constat": "Même ligne (GPT-5 mini, 34,8 %, datée du 24 décembre 2025) affichée sous spoox-m jusqu'au 4 mai 2026, puis sous spoox-o-m à partir du 19 mai 2026.",
   "sources": [
    "https://web.archive.org/web/20260504052926/https://www.tbench.ai/leaderboard/terminal-bench/2.0",
    "https://web.archive.org/web/20260519111531/https://www.tbench.ai/leaderboard/terminal-bench/2.0"
   ]
  }
 ],
 "par_outil": [
  {
   "harnais": "openhands",
   "nom": "OpenHands",
   "resultats": [
    {
     "modele": "claude-opus-4-8",
     "benchmark": "gso_bench",
     "valeur": 47.06,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "claude-opus-4-7",
     "benchmark": "gso_bench",
     "valeur": 44.12,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "claude-opus-4-6",
     "benchmark": "gso_bench",
     "valeur": 41.2,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "gpt-5-5",
     "benchmark": "gso_bench",
     "valeur": 40.2,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "claude-sonnet-5",
     "benchmark": "gso_bench",
     "valeur": 37.25,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "gpt-5-4",
     "benchmark": "gso_bench",
     "valeur": 31.37,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "gpt-5-2",
     "benchmark": "gso_bench",
     "valeur": 27.45,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "claude-opus-4-5",
     "benchmark": "gso_bench",
     "valeur": 26.47,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "gemini-3-1-pro",
     "benchmark": "gso_bench",
     "valeur": 22.55,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "gemini-3-pro",
     "benchmark": "gso_bench",
     "valeur": 18.63,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "claude-sonnet-4-5",
     "benchmark": "gso_bench",
     "valeur": 14.71,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "gpt-5-1",
     "benchmark": "gso_bench",
     "valeur": 13.73,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "gemini-3-flash",
     "benchmark": "gso_bench",
     "valeur": 9.8,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "o3",
     "benchmark": "gso_bench",
     "valeur": 8.8,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "gpt-5",
     "benchmark": "gso_bench",
     "valeur": 6.86,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "claude-opus-4",
     "benchmark": "gso_bench",
     "valeur": 6.86,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "claude-sonnet-4",
     "benchmark": "gso_bench",
     "valeur": 4.9,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "qwen3-coder-480b-a35b",
     "benchmark": "gso_bench",
     "valeur": 4.9,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "kimi-k2-jul-2025",
     "benchmark": "gso_bench",
     "valeur": 4.9,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "claude-3-5-sonnet-october-2024",
     "benchmark": "gso_bench",
     "valeur": 4.6,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "gemini-2-5-pro-jun-2025",
     "benchmark": "gso_bench",
     "valeur": 3.92,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "claude-3-7-sonnet",
     "benchmark": "gso_bench",
     "valeur": 3.8,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "o4-mini",
     "benchmark": "gso_bench",
     "valeur": 3.6,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "glm-4-5-air",
     "benchmark": "gso_bench",
     "valeur": 2.94,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "o3-mini",
     "benchmark": "gso_bench",
     "valeur": 1.3,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "gpt-4o-nov-2024",
     "benchmark": "gso_bench",
     "valeur": 0,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "claude-opus-4-5",
     "benchmark": "terminal_bench",
     "valeur": 51.91,
     "unite": "%",
     "rang": 7,
     "sur": 8
    },
    {
     "modele": "gpt-5",
     "benchmark": "terminal_bench",
     "valeur": 43.82,
     "unite": "%",
     "rang": 2,
     "sur": 4
    },
    {
     "modele": "claude-sonnet-4-5",
     "benchmark": "terminal_bench",
     "valeur": 42.58,
     "unite": "%",
     "rang": 5,
     "sur": 7
    },
    {
     "modele": "claude-opus-4-1",
     "benchmark": "terminal_bench",
     "valeur": 36.85,
     "unite": "%",
     "rang": 2,
     "sur": 4
    },
    {
     "modele": "gpt-5-mini",
     "benchmark": "terminal_bench",
     "valeur": 29.21,
     "unite": "%",
     "rang": 3,
     "sur": 5
    },
    {
     "modele": "grok-4",
     "benchmark": "terminal_bench",
     "valeur": 27.19,
     "unite": "%",
     "rang": 1,
     "sur": 3
    },
    {
     "modele": "kimi-k2-jul-2025",
     "benchmark": "terminal_bench",
     "valeur": 26.74,
     "unite": "%",
     "rang": 2,
     "sur": 2
    },
    {
     "modele": "qwen3-coder-480b-a35b",
     "benchmark": "terminal_bench",
     "valeur": 25.39,
     "unite": "%",
     "rang": 2,
     "sur": 3
    },
    {
     "modele": "gemini-2-5-pro-jun-2025",
     "benchmark": "terminal_bench",
     "valeur": 16.4,
     "unite": "%",
     "rang": 4,
     "sur": 4
    },
    {
     "modele": "gemini-2-5-flash-jun-2025",
     "benchmark": "terminal_bench",
     "valeur": 16.4,
     "unite": "%",
     "rang": 3,
     "sur": 4
    },
    {
     "modele": "gemini-2-5-flash-sep-2025",
     "benchmark": "terminal_bench",
     "valeur": 16.4,
     "unite": "%",
     "rang": 3,
     "sur": 4
    },
    {
     "modele": "claude-haiku-4-5",
     "benchmark": "terminal_bench",
     "valeur": 13.93,
     "unite": "%",
     "rang": 5,
     "sur": 5
    },
    {
     "modele": "gpt-5-nano",
     "benchmark": "terminal_bench",
     "valeur": 9.89,
     "unite": "%",
     "rang": 3,
     "sur": 5
    }
   ]
  },
  {
   "harnais": "claude-code",
   "nom": "Claude Code",
   "resultats": [
    {
     "modele": "claude-opus-5-5",
     "benchmark": "frontiercode",
     "valeur": 54.64,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "claude-fable-5",
     "benchmark": "frontiercode",
     "valeur": 53.48,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "claude-opus-5",
     "benchmark": "frontiercode",
     "valeur": 53.38,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "claude-sonnet-5-5",
     "benchmark": "frontiercode",
     "valeur": 52.09,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "claude-fable-5-1",
     "benchmark": "frontiercode",
     "valeur": 50.91,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "claude-opus-4-8",
     "benchmark": "frontiercode",
     "valeur": 46.5,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "claude-sonnet-5",
     "benchmark": "frontiercode",
     "valeur": 42.73,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "claude-opus-4-7",
     "benchmark": "frontiercode",
     "valeur": 38.54,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "claude-opus-4-6",
     "benchmark": "frontiercode",
     "valeur": 26.64,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "claude-sonnet-4-6",
     "benchmark": "frontiercode",
     "valeur": 24.31,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "claude-fable-5",
     "benchmark": "posttrainbench",
     "valeur": 41.79,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "claude-opus-5",
     "benchmark": "posttrainbench",
     "valeur": 35.04,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "claude-opus-4-8",
     "benchmark": "posttrainbench",
     "valeur": 33.84,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "kimi-k3",
     "benchmark": "posttrainbench",
     "valeur": 31.96,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "glm-5-2",
     "benchmark": "posttrainbench",
     "valeur": 31.7,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "claude-opus-4-7",
     "benchmark": "posttrainbench",
     "valeur": 28.56,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "claude-opus-4-6",
     "benchmark": "terminal_bench",
     "valeur": 57.98,
     "unite": "%",
     "rang": 10,
     "sur": 10
    },
    {
     "modele": "claude-opus-4-5",
     "benchmark": "terminal_bench",
     "valeur": 52.13,
     "unite": "%",
     "rang": 6,
     "sur": 8
    },
    {
     "modele": "claude-sonnet-4-5",
     "benchmark": "terminal_bench",
     "valeur": 40.06,
     "unite": "%",
     "rang": 7,
     "sur": 7
    },
    {
     "modele": "claude-opus-4-1",
     "benchmark": "terminal_bench",
     "valeur": 34.83,
     "unite": "%",
     "rang": 4,
     "sur": 4
    },
    {
     "modele": "claude-haiku-4-5",
     "benchmark": "terminal_bench",
     "valeur": 27.53,
     "unite": "%",
     "rang": 4,
     "sur": 5
    },
    {
     "modele": "claude-opus-5-5",
     "benchmark": "weirdml_v3",
     "valeur": 31.2,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "claude-fable-5-1",
     "benchmark": "weirdml_v3",
     "valeur": 25.97,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "claude-opus-5",
     "benchmark": "weirdml_v3",
     "valeur": 19.01,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "claude-fable-5",
     "benchmark": "weirdml_v3",
     "valeur": 15.53,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "claude-opus-4-5",
     "benchmark": "weirdml_v3",
     "valeur": 6.52,
     "unite": "%",
     "rang": null,
     "sur": null
    }
   ]
  },
  {
   "harnais": "codex-cli",
   "nom": "Codex CLI",
   "resultats": [
    {
     "modele": "gpt-5-5",
     "benchmark": "exploitbench",
     "valeur": 41.6,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "gpt-6-astra",
     "benchmark": "frontiercode",
     "valeur": 53.26,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "gpt-6-1-sol",
     "benchmark": "frontiercode",
     "valeur": 50.23,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "gpt-6-sol",
     "benchmark": "frontiercode",
     "valeur": 49.27,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "gpt-5-6-sol",
     "benchmark": "frontiercode",
     "valeur": 47.49,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "gpt-5-5",
     "benchmark": "frontiercode",
     "valeur": 42.96,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "gpt-6-luna",
     "benchmark": "frontiercode",
     "valeur": 42.42,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "gpt-5-6-terra",
     "benchmark": "frontiercode",
     "valeur": 41.31,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "gpt-5-6-luna",
     "benchmark": "frontiercode",
     "valeur": 39.81,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "gpt-5-4-mini",
     "benchmark": "frontiercode",
     "valeur": 27.04,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "gpt-5-6-sol",
     "benchmark": "posttrainbench",
     "valeur": 36.23,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "gpt-5-5",
     "benchmark": "posttrainbench",
     "valeur": 27.23,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "gpt-5-4",
     "benchmark": "posttrainbench",
     "valeur": 19,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "gpt-5-5",
     "benchmark": "terminal_bench",
     "valeur": 82.25,
     "unite": "%",
     "rang": 3,
     "sur": 4
    },
    {
     "modele": "gpt-5-2",
     "benchmark": "terminal_bench",
     "valeur": 62.92,
     "unite": "%",
     "rang": 2,
     "sur": 4
    },
    {
     "modele": "gpt-5-1-codex-max",
     "benchmark": "terminal_bench",
     "valeur": 60.45,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "gpt-5",
     "benchmark": "terminal_bench",
     "valeur": 49.61,
     "unite": "%",
     "rang": 1,
     "sur": 4
    },
    {
     "modele": "gpt-5-codex",
     "benchmark": "terminal_bench",
     "valeur": 44.33,
     "unite": "%",
     "rang": 1,
     "sur": 3
    },
    {
     "modele": "gpt-5-mini",
     "benchmark": "terminal_bench",
     "valeur": 31.85,
     "unite": "%",
     "rang": 2,
     "sur": 5
    },
    {
     "modele": "gpt-5-nano",
     "benchmark": "terminal_bench",
     "valeur": 11.46,
     "unite": "%",
     "rang": 2,
     "sur": 5
    },
    {
     "modele": "gpt-6-astra",
     "benchmark": "weirdml_v3",
     "valeur": 42.22,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "gpt-6-sol",
     "benchmark": "weirdml_v3",
     "valeur": 19.73,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "gpt-5-6-sol",
     "benchmark": "weirdml_v3",
     "valeur": 14.99,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "gpt-6-luna",
     "benchmark": "weirdml_v3",
     "valeur": 7.87,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "gpt-5-6-luna",
     "benchmark": "weirdml_v3",
     "valeur": 5.16,
     "unite": "%",
     "rang": null,
     "sur": null
    }
   ]
  },
  {
   "harnais": "gemini-cli",
   "nom": "Gemini CLI",
   "resultats": [
    {
     "modele": "gemini-3-1-pro",
     "benchmark": "terminal_bench",
     "valeur": 59.36,
     "unite": "%",
     "rang": 3,
     "sur": 3
    },
    {
     "modele": "gemini-3-flash",
     "benchmark": "terminal_bench",
     "valeur": 47.42,
     "unite": "%",
     "rang": 3,
     "sur": 3
    },
    {
     "modele": "gemini-2-5-pro-jun-2025",
     "benchmark": "terminal_bench",
     "valeur": 19.55,
     "unite": "%",
     "rang": 3,
     "sur": 4
    },
    {
     "modele": "gemini-2-5-flash-jun-2025",
     "benchmark": "terminal_bench",
     "valeur": 15.45,
     "unite": "%",
     "rang": 4,
     "sur": 4
    },
    {
     "modele": "gemini-2-5-flash-sep-2025",
     "benchmark": "terminal_bench",
     "valeur": 15.4,
     "unite": "%",
     "rang": 4,
     "sur": 4
    },
    {
     "modele": "gemini-3-8-flash",
     "benchmark": "weirdml_v3",
     "valeur": 7.38,
     "unite": "%",
     "rang": null,
     "sur": null
    }
   ]
  },
  {
   "harnais": "droid",
   "nom": "Droid (Factory)",
   "resultats": [
    {
     "modele": "gpt-5-3-codex",
     "benchmark": "terminal_bench",
     "valeur": 77.3,
     "unite": "%",
     "rang": 2,
     "sur": 9
    },
    {
     "modele": "claude-opus-4-6",
     "benchmark": "terminal_bench",
     "valeur": 69.89,
     "unite": "%",
     "rang": 6,
     "sur": 10
    },
    {
     "modele": "gpt-5-2",
     "benchmark": "terminal_bench",
     "valeur": 64.94,
     "unite": "%",
     "rang": 1,
     "sur": 4
    },
    {
     "modele": "claude-opus-4-5",
     "benchmark": "terminal_bench",
     "valeur": 63.15,
     "unite": "%",
     "rang": 1,
     "sur": 8
    },
    {
     "modele": "gemini-3-pro",
     "benchmark": "terminal_bench",
     "valeur": 61.12,
     "unite": "%",
     "rang": 5,
     "sur": 7
    }
   ]
  },
  {
   "harnais": "mux",
   "nom": "Xum (ex-Mux)",
   "resultats": [
    {
     "modele": "gpt-5-3-codex",
     "benchmark": "terminal_bench",
     "valeur": 74.61,
     "unite": "%",
     "rang": 6,
     "sur": 9
    },
    {
     "modele": "claude-opus-4-6",
     "benchmark": "terminal_bench",
     "valeur": 66.52,
     "unite": "%",
     "rang": 8,
     "sur": 10
    },
    {
     "modele": "gpt-5-2",
     "benchmark": "terminal_bench",
     "valeur": 60.67,
     "unite": "%",
     "rang": 3,
     "sur": 4
    },
    {
     "modele": "claude-opus-4-5",
     "benchmark": "terminal_bench",
     "valeur": 58.43,
     "unite": "%",
     "rang": 3,
     "sur": 8
    }
   ]
  },
  {
   "harnais": "goose",
   "nom": "goose",
   "resultats": [
    {
     "modele": "claude-opus-4-5",
     "benchmark": "terminal_bench",
     "valeur": 54.33,
     "unite": "%",
     "rang": 5,
     "sur": 8
    },
    {
     "modele": "claude-sonnet-4-5",
     "benchmark": "terminal_bench",
     "valeur": 43.15,
     "unite": "%",
     "rang": 2,
     "sur": 7
    },
    {
     "modele": "claude-haiku-4-5",
     "benchmark": "terminal_bench",
     "valeur": 35.51,
     "unite": "%",
     "rang": 1,
     "sur": 5
    }
   ]
  },
  {
   "harnais": "letta-code",
   "nom": "Letta Code",
   "resultats": [
    {
     "modele": "claude-opus-4-5",
     "benchmark": "terminal_bench",
     "valeur": 59.1,
     "unite": "%",
     "rang": 2,
     "sur": 8
    },
    {
     "modele": "gemini-3-pro",
     "benchmark": "terminal_bench",
     "valeur": 55.96,
     "unite": "%",
     "rang": 7,
     "sur": 7
    },
    {
     "modele": "gpt-5-1-codex",
     "benchmark": "terminal_bench",
     "valeur": 53.48,
     "unite": "%",
     "rang": 2,
     "sur": 3
    }
   ]
  },
  {
   "harnais": "opencode",
   "nom": "OpenCode",
   "resultats": [
    {
     "modele": "gemini-3-1-pro",
     "benchmark": "posttrainbench",
     "valeur": 21.99,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "claude-opus-4-5",
     "benchmark": "terminal_bench",
     "valeur": 51.69,
     "unite": "%",
     "rang": 8,
     "sur": 8
    },
    {
     "modele": "deepseek-v4-1-flash",
     "benchmark": "weirdml_v3",
     "valeur": 5.95,
     "unite": "%",
     "rang": null,
     "sur": null
    }
   ]
  },
  {
   "harnais": "cursor-cli",
   "nom": "Cursor CLI",
   "resultats": [
    {
     "modele": "composer-2-5",
     "benchmark": "frontiercode",
     "valeur": 25.64,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "grok-4-5",
     "benchmark": "posttrainbench",
     "valeur": 23.45,
     "unite": "%",
     "rang": null,
     "sur": null
    }
   ]
  },
  {
   "harnais": "grok-build",
   "nom": "Grok Build",
   "resultats": [
    {
     "modele": "grok-4-6",
     "benchmark": "frontiercode",
     "valeur": 48.01,
     "unite": "%",
     "rang": null,
     "sur": null
    },
    {
     "modele": "grok-4-5",
     "benchmark": "frontiercode",
     "valeur": 42.44,
     "unite": "%",
     "rang": null,
     "sur": null
    }
   ]
  },
  {
   "harnais": "devin",
   "nom": "Devin",
   "resultats": [
    {
     "modele": "swe-2",
     "benchmark": "frontiercode",
     "valeur": 50,
     "unite": "%",
     "rang": null,
     "sur": null
    }
   ]
  },
  {
   "harnais": "grok-cli",
   "nom": "Grok CLI",
   "resultats": [
    {
     "modele": "grok-4-20",
     "benchmark": "terminal_bench",
     "valeur": 57.3,
     "unite": "%",
     "rang": null,
     "sur": null
    }
   ]
  },
  {
   "harnais": "junie-cli",
   "nom": "Junie CLI",
   "resultats": [
    {
     "modele": "gemini-3-flash",
     "benchmark": "terminal_bench",
     "valeur": 64.27,
     "unite": "%",
     "rang": 1,
     "sur": 3
    }
   ]
  },
  {
   "harnais": "kimi-code",
   "nom": "Kimi Code CLI",
   "resultats": [
    {
     "modele": "kimi-k3",
     "benchmark": "weirdml_v3",
     "valeur": 7.27,
     "unite": "%",
     "rang": null,
     "sur": null
    }
   ]
  }
 ]
}