{"source":"Quelle IA, le classement des modèles de langage","page":"https://quelleia.com/benchmarks/scienceqa/","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":"scienceqa","nom":"ScienceQA","tache":"autre","etat":"archive","echelle":"unite","modeles":23,"poids_general":0,"difficulte":4.3,"pente":0.0455,"en_une_phrase":"Des questions de sciences de niveau école et collège, à choix multiples, souvent accompagnées d'une image ou d'un schéma.","notation":"Part de bonnes réponses à choix multiples.","limites":"La plupart des résultats concernent des modèles de recherche entraînés spécialement sur ce jeu, peu comparables aux assistants grand public.","mainteneur":"Lu et al. (2022)","url":"https://scienceqa.github.io/","verifie":true},"tache":{"id":"autre","nom":"Tests historiques","poids_general":0,"benchmarks_actifs":0},"modeles_mesures":23,"mesures":26,"classement":[{"rang":1,"modele":"phi-3-5-vision-instruct","nom":"Phi-3.5-vision-instruct","editeur":"Microsoft","page":"https://quelleia.com/modeles/phi-3-5-vision-instruct/","score_publie":0.913,"score_0_1":0.884,"niveau_suggere":48.9,"reflexion":"inconnue","harnais":null,"configurations_mesurees":1,"date":null,"source":"https://epoch.ai/benchmarks"},{"rang":2,"modele":"gpt-4o-may-2024","nom":"GPT-4o (May 2024)","editeur":"OpenAI","page":"https://quelleia.com/modeles/gpt-4o-may-2024/","score_publie":0.885,"score_0_1":0.8467,"niveau_suggere":41.8,"reflexion":"inconnue","harnais":null,"configurations_mesurees":1,"date":null,"source":"https://epoch.ai/benchmarks"},{"rang":3,"modele":"gemini-1-0-pro-vision","nom":"Gemini 1.0 Pro Vision","editeur":"Google DeepMind","page":"https://quelleia.com/modeles/gemini-1-0-pro-vision/","score_publie":0.797,"score_0_1":0.7293,"niveau_suggere":26,"reflexion":"inconnue","harnais":null,"configurations_mesurees":1,"date":null,"source":"https://epoch.ai/benchmarks"},{"rang":4,"modele":"falcon-11b-vlm","nom":"falcon-11B-vlm","editeur":"Technology Innovation Institute","page":"https://quelleia.com/modeles/falcon-11b-vlm/","score_publie":0.749,"score_0_1":0.6653,"niveau_suggere":19.4,"reflexion":"inconnue","harnais":null,"configurations_mesurees":1,"date":null,"source":"https://epoch.ai/benchmarks"},{"rang":5,"modele":"blip-2-q-former","nom":"BLIP-2 (Q-Former)","editeur":"Salesforce Research","page":"https://quelleia.com/modeles/blip-2-q-former/","score_publie":0.7417,"score_0_1":0.6556,"niveau_suggere":18.4,"reflexion":"inconnue","harnais":null,"configurations_mesurees":1,"date":null,"source":"https://epoch.ai/benchmarks"},{"rang":6,"modele":"instructgpt-175b","nom":"InstructGPT 175B","editeur":"OpenAI","page":"https://quelleia.com/modeles/instructgpt-175b/","score_publie":0.7404,"score_0_1":0.6539,"niveau_suggere":18.2,"reflexion":"inconnue","harnais":null,"configurations_mesurees":2,"date":null,"source":"https://epoch.ai/benchmarks"},{"rang":7,"modele":"llama3-llava-next-8b","nom":"llama3-llava-next-8b","editeur":null,"page":"https://quelleia.com/modeles/llama3-llava-next-8b/","score_publie":0.737,"score_0_1":0.6493,"niveau_suggere":17.8,"reflexion":"inconnue","harnais":null,"configurations_mesurees":1,"date":null,"source":"https://epoch.ai/benchmarks"},{"rang":8,"modele":"llava-v1-6-vicuna-13b","nom":"llava-v1.6-vicuna-13b","editeur":null,"page":"https://quelleia.com/modeles/llava-v1-6-vicuna-13b/","score_publie":0.736,"score_0_1":0.648,"niveau_suggere":17.7,"reflexion":"inconnue","harnais":null,"configurations_mesurees":1,"date":null,"source":"https://epoch.ai/benchmarks"},{"rang":9,"modele":"llava-v1-6-mistral-7b","nom":"llava-v1.6-mistral-7b","editeur":null,"page":"https://quelleia.com/modeles/llava-v1-6-mistral-7b/","score_publie":0.728,"score_0_1":0.6373,"niveau_suggere":16.7,"reflexion":"inconnue","harnais":null,"configurations_mesurees":1,"date":null,"source":"https://epoch.ai/benchmarks"},{"rang":10,"modele":"mm1-7b-chat","nom":"MM1-7B-Chat","editeur":"Apple","page":"https://quelleia.com/modeles/mm1-7b-chat/","score_publie":0.726,"score_0_1":0.6347,"niveau_suggere":16.4,"reflexion":"inconnue","harnais":null,"configurations_mesurees":1,"date":null,"source":"https://epoch.ai/benchmarks"},{"rang":11,"modele":"claude-3-haiku","nom":"Claude 3 Haiku","editeur":"Anthropic","page":"https://quelleia.com/modeles/claude-3-haiku/","score_publie":0.72,"score_0_1":0.6267,"niveau_suggere":15.6,"reflexion":"inconnue","harnais":null,"configurations_mesurees":1,"date":null,"source":"https://epoch.ai/benchmarks"},{"rang":12,"modele":"llava-v1-6-vicuna-7b","nom":"llava-v1.6-vicuna-7b","editeur":null,"page":"https://quelleia.com/modeles/llava-v1-6-vicuna-7b/","score_publie":0.706,"score_0_1":0.608,"niveau_suggere":13.9,"reflexion":"inconnue","harnais":null,"configurations_mesurees":2,"date":null,"source":"https://epoch.ai/benchmarks"},{"rang":13,"modele":"internvl-chat-vit-6b-vicuna-13b","nom":"InternVL-Chat-ViT-6B-Vicuna-13B","editeur":"Shanghai AI Laboratory","page":"https://quelleia.com/modeles/internvl-chat-vit-6b-vicuna-13b/","score_publie":0.701,"score_0_1":0.6013,"niveau_suggere":13.3,"reflexion":"inconnue","harnais":null,"configurations_mesurees":1,"date":null,"source":"https://epoch.ai/benchmarks"},{"rang":14,"modele":"mm1-3b-chat","nom":"MM1-3B-Chat","editeur":"Apple","page":"https://quelleia.com/modeles/mm1-3b-chat/","score_publie":0.694,"score_0_1":0.592,"niveau_suggere":12.4,"reflexion":"inconnue","harnais":null,"configurations_mesurees":1,"date":null,"source":"https://epoch.ai/benchmarks"},{"rang":15,"modele":"qwen-vl-chat","nom":"Qwen-VL-Chat","editeur":"Alibaba","page":"https://quelleia.com/modeles/qwen-vl-chat/","score_publie":0.682,"score_0_1":0.576,"niveau_suggere":11,"reflexion":"inconnue","harnais":null,"configurations_mesurees":2,"date":null,"source":"https://epoch.ai/benchmarks"},{"rang":16,"modele":"llava-v1-5-7b","nom":"llava-v1.5-7b","editeur":null,"page":"https://quelleia.com/modeles/llava-v1-5-7b/","score_publie":0.668,"score_0_1":0.5573,"niveau_suggere":9.3,"reflexion":"inconnue","harnais":null,"configurations_mesurees":1,"date":null,"source":"https://epoch.ai/benchmarks"},{"rang":17,"modele":"internvl-chat-vit-6b-vicuna-7b","nom":"InternVL-Chat-ViT-6B-Vicuna-7B","editeur":"Shanghai AI Laboratory","page":"https://quelleia.com/modeles/internvl-chat-vit-6b-vicuna-7b/","score_publie":0.662,"score_0_1":0.5493,"niveau_suggere":8.6,"reflexion":"inconnue","harnais":null,"configurations_mesurees":1,"date":null,"source":"https://epoch.ai/benchmarks"},{"rang":18,"modele":"instructblip-vicuna-13b","nom":"instructblip-vicuna-13b","editeur":null,"page":"https://quelleia.com/modeles/instructblip-vicuna-13b/","score_publie":0.631,"score_0_1":0.508,"niveau_suggere":5,"reflexion":"inconnue","harnais":null,"configurations_mesurees":1,"date":null,"source":"https://epoch.ai/benchmarks"},{"rang":19,"modele":"instructblip-vicuna-7b","nom":"instructblip-vicuna-7b","editeur":null,"page":"https://quelleia.com/modeles/instructblip-vicuna-7b/","score_publie":0.605,"score_0_1":0.4733,"niveau_suggere":1.9,"reflexion":"inconnue","harnais":null,"configurations_mesurees":1,"date":null,"source":"https://epoch.ai/benchmarks"},{"rang":20,"modele":"llama-2-13b","nom":"Llama 2-13B","editeur":"Meta AI","page":"https://quelleia.com/modeles/llama-2-13b/","score_publie":0.5578,"score_0_1":0.4104,"niveau_suggere":-3.7,"reflexion":"inconnue","harnais":null,"configurations_mesurees":1,"date":null,"source":"https://epoch.ai/benchmarks"},{"rang":21,"modele":"llama-13b","nom":"LLaMA-13B","editeur":"Meta AI","page":"https://quelleia.com/modeles/llama-13b/","score_publie":0.4333,"score_0_1":0.2444,"niveau_suggere":-20.6,"reflexion":"inconnue","harnais":null,"configurations_mesurees":1,"date":null,"source":"https://epoch.ai/benchmarks"},{"rang":22,"modele":"llama-2-7b","nom":"Llama 2-7B","editeur":"Meta AI","page":"https://quelleia.com/modeles/llama-2-7b/","score_publie":0.4308,"score_0_1":0.2411,"niveau_suggere":-21,"reflexion":"inconnue","harnais":null,"configurations_mesurees":1,"date":null,"source":"https://epoch.ai/benchmarks"},{"rang":23,"modele":"llama-7b","nom":"LLaMA-7B","editeur":"Meta AI","page":"https://quelleia.com/modeles/llama-7b/","score_publie":0.3619,"score_0_1":0.1492,"niveau_suggere":-34,"reflexion":"inconnue","harnais":null,"configurations_mesurees":1,"date":null,"source":"https://epoch.ai/benchmarks"}]}