{"source":"Quelle IA, le classement des modèles de langage","page":"https://quelleia.com/benchmarks/exploitbench/","edition":"2026-W40","date":"2026-09-28","echelle":"score_publie est le score de la source, dans son unité. score_0_1 est ce score ramené entre 0 et 1 pour le calcul. niveau_suggere est le score IA que cette seule mesure indique : 100 pour le meilleur modèle au lancement du site, 50 pour GPT-4o de novembre 2024. difficulte est le score IA d'un modèle qui obtient la moitié (score_0_1 = 0,5), pente la raideur de la courbe.","licence":"Réutilisation libre en citant « Quelle IA, quelleia.com » et la date de l'édition. Les mesures renvoient à leurs sources d'origine.","benchmark":{"id":"exploitbench","nom":"ExploitBench","tache":"securite","etat":"actif","echelle":"unite","modeles":9,"poids_general":0,"difficulte":95.3,"pente":0.1041,"en_une_phrase":"Exploiter de vraies failles déjà corrigées du moteur JavaScript de Chrome, de l'accès au code vulnérable jusqu'à l'exécution de code arbitraire.","notation":"Part moyenne des 16 jalons atteints sur 41 failles, avec plusieurs essais par faille.","limites":"Un seul logiciel cible, le moteur V8. Peu de modèles testés, certains en préversion, et les failles sont déjà documentées publiquement.","mainteneur":"Université Carnegie Mellon","url":"https://exploitbench.ai/","verifie":true},"tache":{"id":"securite","nom":"Sécurité","poids_general":0,"benchmarks_actifs":0},"modeles_mesures":9,"mesures":20,"classement":[{"rang":1,"modele":"claude-mythos-preview","nom":"Claude Mythos Preview","editeur":"Anthropic","page":"https://quelleia.com/modeles/claude-mythos-preview/","score_publie":0.738,"score_0_1":0.738,"niveau_suggere":105.3,"reflexion":"inconnue","harnais":null,"configurations_mesurees":2,"date":null,"source":"https://epoch.ai/benchmarks"},{"rang":2,"modele":"gpt-5-5","nom":"GPT-5.5","editeur":"OpenAI","page":"https://quelleia.com/modeles/gpt-5-5/","score_publie":0.474,"score_0_1":0.474,"niveau_suggere":94.3,"reflexion":"inconnue","harnais":null,"configurations_mesurees":4,"date":null,"source":"https://epoch.ai/benchmarks"},{"rang":3,"modele":"claude-opus-4-7","nom":"Claude Opus 4.7","editeur":"Anthropic","page":"https://quelleia.com/modeles/claude-opus-4-7/","score_publie":0.265,"score_0_1":0.265,"niveau_suggere":85.5,"reflexion":"inconnue","harnais":null,"configurations_mesurees":2,"date":null,"source":"https://epoch.ai/benchmarks"},{"rang":4,"modele":"gemini-3-1-pro","nom":"Gemini 3.1 Pro","editeur":"Google DeepMind","page":"https://quelleia.com/modeles/gemini-3-1-pro/","score_publie":0.261,"score_0_1":0.261,"niveau_suggere":85.3,"reflexion":"inconnue","harnais":null,"configurations_mesurees":2,"date":null,"source":"https://epoch.ai/benchmarks"},{"rang":5,"modele":"claude-sonnet-4-6","nom":"Claude Sonnet 4.6","editeur":"Anthropic","page":"https://quelleia.com/modeles/claude-sonnet-4-6/","score_publie":0.236,"score_0_1":0.236,"niveau_suggere":84,"reflexion":"inconnue","harnais":null,"configurations_mesurees":2,"date":null,"source":"https://epoch.ai/benchmarks"},{"rang":6,"modele":"kimi-k2-6","nom":"Kimi K2.6","editeur":"Moonshot","page":"https://quelleia.com/modeles/kimi-k2-6/","score_publie":0.184,"score_0_1":0.184,"niveau_suggere":81,"reflexion":"inconnue","harnais":null,"configurations_mesurees":2,"date":null,"source":"https://epoch.ai/benchmarks"},{"rang":7,"modele":"glm-5-1","nom":"GLM-5.1","editeur":"Z.ai (Zhipu AI)","page":"https://quelleia.com/modeles/glm-5-1/","score_publie":0.181,"score_0_1":0.181,"niveau_suggere":80.8,"reflexion":"inconnue","harnais":null,"configurations_mesurees":2,"date":null,"source":"https://epoch.ai/benchmarks"},{"rang":8,"modele":"claude-haiku-4-5","nom":"Claude Haiku 4.5","editeur":"Anthropic","page":"https://quelleia.com/modeles/claude-haiku-4-5/","score_publie":0.1369999999999999,"score_0_1":0.137,"niveau_suggere":77.6,"reflexion":"inconnue","harnais":null,"configurations_mesurees":2,"date":null,"source":"https://epoch.ai/benchmarks"},{"rang":9,"modele":"minimax-m2-7","nom":"MiniMax-M2.7","editeur":"MiniMax","page":"https://quelleia.com/modeles/minimax-m2-7/","score_publie":0.133,"score_0_1":0.133,"niveau_suggere":77.3,"reflexion":"inconnue","harnais":null,"configurations_mesurees":2,"date":null,"source":"https://epoch.ai/benchmarks"}]}