Aidez à maintenir l’exactitude des données publiques
Une information manque, est incorrecte ou obsolète ? Signalez-le-nous pour cette page.
Une information manque, est incorrecte ou obsolète ? Signalez-le-nous pour cette page.
Résultats des évaluations de l’IA et performances des modèles
Scores individuels du benchmark tracés par date.
| Organisation | Modèles | Date du résultat | Meilleur score | Informations | Déclaré par le fournisseur | Source |
|---|---|---|---|---|---|---|
| Gemini 4 Argon | 13 oct. 2026 | 77,90 % | Google-reported launch evaluation; source chart: https://storage.googleapis.com/gweb-uniblog-publish-prod/original_images/gemini-4-argon_table_blog.gif | Oui | Source | |
| DeepSeek V4.1 Flash | 10 sept. 2026 | 74,20 % | Instruct model; reasoning_effort=100; mini-SWE harness; Resolved | Oui | Source | |
| GPT-6 Astra | 03 sept. 2026 | 74,10 % | DeepSWE v1.1 | Oui | Source | |
| GPT 5.5 | 23 avr. 2026 | 70 % | mini-swe-agent; effort xhigh; 70% +/-3%; avg cost $6.61; avg time 21m; out tok 47k | Non | Source | |
| Claude Opus 5 | 24 juil. 2026 | 68,80 % | DeepSWE v1.1; 113 long-horizon tasks; adaptive thinking at max effort; average of 5 trials | Oui | Source | |
| Kimi K3 | 16 juil. 2026 | 67,50 % | KimiCode harness; Moonshot release-table score. The official DeepSWE leaderboard reports a separate 67.3 result with mini-SWE-agent. | Oui | Source | |
| Claude Opus 4.8 | 28 mai 2026 | 58 % | mini-swe-agent; effort max; 58% +/-2%; avg cost $12.58; avg time 43m; out tok 136k | Non | Source | |
| GPT 5.4 | 05 mars 2026 | 56 % | mini-swe-agent; effort xhigh; 56% +/-2%; avg cost $4.38; avg time 27m; out tok 71k | Non | Source | |
| DeepSeek V4 Flash 0731 | 31 juil. 2026 | 54,40 % | DeepSeek-V4-Flash-0731 | Oui | Source | |
| Claude Opus 4.7 | 16 avr. 2026 | 54 % | mini-swe-agent; effort max; 54% +/-5%; avg cost $18.19; avg time 39m; out tok 103k | Non | Source | |
| Muse Spark 1.1 | 09 juil. 2026 | 53,30 % | DeepSWE 1.1; long-horizon agentic coding | Oui | Source | |
| Qwen 3.8 27B | 14 août 2026 | 42,20 % | DeepSWE 1.1 | Oui | Source | |
| Claude Sonnet 4.6 | 17 févr. 2026 | 32 % | mini-swe-agent; effort high; 32% +/-2%; avg cost $5.52; avg time 42m; out tok 76k | Non | Source | |
| Claude Opus 4.6 | 05 févr. 2026 | 28 % | mini-swe-agent; effort max; 28% +/-4%; avg cost $5.39; avg time 30m; out tok 44k | Non | Source | |
| Gemini 3.5 Flash | 19 mai 2026 | 28 % | mini-swe-agent; effort medium; 28% +/-4%; avg cost $7.42; avg time 17m; out tok 189k | Non | Source | |
| Kimi K2.6 | 20 avr. 2026 | 24 % | mini-swe-agent; 24% +/-2%; avg cost $3.16; avg time 56m; out tok 84k | Non | Source | |
| GPT 5.4 Mini | 17 mars 2026 | 24 % | mini-swe-agent; effort xhigh; 24% +/-3%; avg cost $2.08; avg time 33m; out tok 135k | Non | Source | |
| MiniMax M3 | 01 juin 2026 | 20 % | mini-swe-agent; 20% +/-4%; avg cost $5.57; avg time 57m; out tok 98k | Non | Source | |
| MiMo V2.5 Pro | 23 avr. 2026 | 19 % | mini-swe-agent; 19% +/-2%; avg cost $1.99; avg time 28m; out tok 49k | Non | Source | |
| GLM 5.1 | 07 avr. 2026 | 18 % | mini-swe-agent; 18% +/-1%; avg cost $7.46; avg time 35m; out tok 49k | Non | Source | |
| Grok Build 0.1 | 15 mai 2026 | 13 % | mini-swe-agent; 13% +/-2%; avg cost $6.60; avg time 44m; out tok 52k | Non | Source | |
| Gemini 3.1 Pro Preview | 19 févr. 2026 | 10 % | DeepSWE label: gemini-3.1-pro; mini-swe-agent; 10% +/-3%; avg cost $1.84; avg time 36m; out tok 53k | Non | Source | |
| DeepSeek V4 Pro | 24 avr. 2026 | 8 % | mini-swe-agent; 8% +/-3%; avg cost $4.22; avg time 37m; out tok 50k | Non | Source | |
| Gemini 3 Flash Preview | 17 déc. 2025 | 5 % | DeepSWE label: gemini-3-flash; mini-swe-agent; 5% +/-2%; avg cost $1.53; avg time 39m; out tok 233k | Non | Source |
