{"source":"Quelle IA, le classement des modèles de langage","page":"https://quelleia.com/benchmarks/mirrorcode/","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":"mirrorcode","nom":"MirrorCode","tache":"code","etat":"actif","echelle":"unite","modeles":9,"poids_general":1.15,"difficulte":96,"pente":0.224,"en_une_phrase":"Réécrire de zéro un programme complet (utilitaire Unix, interpréteur, compresseur) sans voir son code, en reproduisant exactement son comportement.","notation":"Part des programmes reproduits à l'identique, tous les tests passés, sur 30 tâches avec trois essais chacune.","limites":"Les programmes d'origine sont publics et ont pu être vus à l'entraînement. Le budget accordé est énorme : jusqu'à 10 milliards de jetons et 7 jours par tâche.","mainteneur":"Epoch AI et METR","url":"https://epoch.ai/benchmarks/mirrorcode","verifie":true},"tache":{"id":"code","nom":"Code","poids_general":15,"benchmarks_actifs":13},"modeles_mesures":9,"mesures":9,"classement":[{"rang":1,"modele":"claude-opus-5-5","nom":"Claude Opus 5.5","editeur":"Anthropic","page":"https://quelleia.com/modeles/claude-opus-5-5/","score_publie":0.774,"score_0_1":0.774,"niveau_suggere":101.5,"reflexion":"maximale","harnais":null,"configurations_mesurees":1,"date":"2026-09-22","source":"https://epoch.ai/benchmarks"},{"rang":2,"modele":"claude-fable-5-1","nom":"Claude Fable 5.1","editeur":"Anthropic","page":"https://quelleia.com/modeles/claude-fable-5-1/","score_publie":0.733,"score_0_1":0.733,"niveau_suggere":100.5,"reflexion":"elevee","harnais":null,"configurations_mesurees":1,"date":"2026-09-10","source":"https://epoch.ai/benchmarks"},{"rang":3,"modele":"claude-fable-5","nom":"Claude Fable 5","editeur":"Anthropic","page":"https://quelleia.com/modeles/claude-fable-5/","score_publie":0.638888888888889,"score_0_1":0.6389,"niveau_suggere":98.6,"reflexion":"elevee","harnais":null,"configurations_mesurees":1,"date":"2026-08-10","source":"https://epoch.ai/benchmarks"},{"rang":4,"modele":"gpt-6-astra","nom":"GPT-6 Astra","editeur":"OpenAI","page":"https://quelleia.com/modeles/gpt-6-astra/","score_publie":0.467,"score_0_1":0.467,"niveau_suggere":95.4,"reflexion":"elevee","harnais":null,"configurations_mesurees":1,"date":"2026-08-30","source":"https://epoch.ai/benchmarks"},{"rang":5,"modele":"claude-opus-4-7","nom":"Claude Opus 4.7","editeur":"Anthropic","page":"https://quelleia.com/modeles/claude-opus-4-7/","score_publie":0.3111111111111111,"score_0_1":0.3111,"niveau_suggere":92.5,"reflexion":"elevee","harnais":null,"configurations_mesurees":1,"date":"2026-08-12","source":"https://epoch.ai/benchmarks"},{"rang":6,"modele":"gpt-5-6-sol","nom":"GPT-5.6 Sol","editeur":"OpenAI","page":"https://quelleia.com/modeles/gpt-5-6-sol/","score_publie":0.2,"score_0_1":0.2,"niveau_suggere":89.8,"reflexion":"elevee","harnais":null,"configurations_mesurees":1,"date":"2026-08-10","source":"https://epoch.ai/benchmarks"},{"rang":7,"modele":"gpt-5-4","nom":"GPT-5.4","editeur":"OpenAI","page":"https://quelleia.com/modeles/gpt-5-4/","score_publie":0.1555555555555555,"score_0_1":0.1556,"niveau_suggere":88.5,"reflexion":"elevee","harnais":null,"configurations_mesurees":1,"date":"2026-08-10","source":"https://epoch.ai/benchmarks"},{"rang":8,"modele":"gpt-5-5","nom":"GPT-5.5","editeur":"OpenAI","page":"https://quelleia.com/modeles/gpt-5-5/","score_publie":0.1,"score_0_1":0.1,"niveau_suggere":86.2,"reflexion":"elevee","harnais":null,"configurations_mesurees":1,"date":"2026-08-10","source":"https://epoch.ai/benchmarks"},{"rang":9,"modele":"gemini-3-1-pro","nom":"Gemini 3.1 Pro","editeur":"Google DeepMind","page":"https://quelleia.com/modeles/gemini-3-1-pro/","score_publie":0.0888888888888888,"score_0_1":0.0889,"niveau_suggere":85.6,"reflexion":"elevee","harnais":null,"configurations_mesurees":1,"date":"2026-08-10","source":"https://epoch.ai/benchmarks"}]}