{"source":"Quelle IA, le classement des modèles de langage","page":"https://quelleia.com/modeles/gpt-2-1-5b/","edition":"2026-W40","date":"2026-09-28","echelle":"Score IA : 100 pour le meilleur modèle au lancement du site, 50 pour GPT-4o de novembre 2024. bas et haut bornent l'intervalle à 90 %.","licence":"Réutilisation libre en citant « Quelle IA, quelleia.com » et la date de l'édition. Les mesures renvoient à leurs sources d'origine.","modele":{"liens":[{"type":"essayer","libelle":"Application de l'éditeur : ChatGPT","url":"https://chatgpt.com"},{"type":"ollama","libelle":"Chercher sur Ollama","url":"https://ollama.com/search?q=GPT-2"},{"type":"versions_allegees","libelle":"Chercher les versions allégées sur Hugging Face","url":"https://huggingface.co/models?library=gguf&search=GPT-2"},{"type":"api","libelle":"Documentation de l'API (OpenAI)","url":"https://platform.openai.com/docs"},{"type":"editeur","libelle":"Site de l'éditeur (OpenAI)","url":"https://openai.com"}],"id":"gpt-2-1-5b","nom":"GPT-2 (1.5B)","editeur":"OpenAI","pays":"États-Unis","date_sortie":"2019-11-05","nouveau":false,"poids_ouverts":true,"acces":"Open weights (unrestricted)","licence":null,"local":null,"prix_eur_par_million":null,"hebergeable_ue":null,"classe_energie":null,"contexte_jetons":null,"entrees":null,"niveaux_reflexion":null,"connaissances_jusqu_a":null,"general":{"score":18.4,"bas":0,"haut":39,"mesures":1,"poids_couvert":15,"sources":1,"provisoire":true,"historique":false},"taches":{"agents":{"score":16.9,"bas":9.2,"haut":24.5,"mesures":1,"benchmarks_tache":9,"couverture":11,"provisoire":true}},"reflexions_mesurees":[],"harnais_mesures":[],"mesures":[{"benchmark":"piqa","tache":"autre","etat":"archive","score_publie":0.705,"score_0_1":0.41,"reflexion":"inconnue","harnais":null,"provenance":"externe","date":null,"source":"https://epoch.ai/benchmarks","niveau_suggere":-95.6},{"benchmark":"boolq","tache":"autre","etat":"archive","score_publie":0.618,"score_0_1":0.236,"reflexion":"inconnue","harnais":null,"provenance":"externe","date":null,"source":"https://epoch.ai/benchmarks","niveau_suggere":-74.1},{"benchmark":"hellaswag","tache":"autre","etat":"archive","score_publie":0.4,"score_0_1":0.2,"reflexion":"inconnue","harnais":null,"provenance":"externe","date":null,"source":"https://epoch.ai/benchmarks","niveau_suggere":-143.3},{"benchmark":"winogrande","tache":"autre","etat":"archive","score_publie":0.583,"score_0_1":0.166,"reflexion":"inconnue","harnais":null,"provenance":"externe","date":null,"source":"https://epoch.ai/benchmarks","niveau_suggere":-72.6},{"benchmark":"winogrande","tache":"autre","etat":"archive","score_publie":0.583,"score_0_1":0.166,"reflexion":"inconnue","harnais":null,"provenance":"externe","date":null,"source":"https://epoch.ai/benchmarks","niveau_suggere":-72.6},{"benchmark":"metr_time_horizons","tache":"agents","etat":"actif","score_publie":0.101046,"score_0_1":0.101,"reflexion":"inconnue","harnais":null,"provenance":"externe","date":null,"source":"https://metr.org/blog/2025-03-19-measuring-ai-ability-to-complete-long-tasks/","niveau_suggere":13.2},{"benchmark":"openbookqa","tache":"autre","etat":"archive","score_publie":0.224,"score_0_1":0,"reflexion":"inconnue","harnais":null,"provenance":"externe","date":null,"source":"https://epoch.ai/benchmarks","niveau_suggere":-135.4},{"benchmark":"arc_ai2","tache":"autre","etat":"archive","score_publie":0.25,"score_0_1":0,"reflexion":"inconnue","harnais":null,"provenance":"externe","date":null,"source":"https://epoch.ai/benchmarks","niveau_suggere":-117.3}]}}