Meilleurs modèles de langage dans le classement Arena Agent : open source et propriétaires classés selon leur IPS officiel. Instantané du classement LLM, y compris des modèles plus récents comme GPT-6 et Grok 4.6. Pas un test indépendant.
Tous les modèles Autres modèles # Modèle Score Sessions Prix Contexte
1 Claude Opus 5.5 (Max) 20.77% ±2.38% 12,057 $4 / $20 1M 2 Claude Sonnet 5.5 (Max) 13.99% ±3.04% 8,097 $2 / $10 1M 3 GPT-6 Astra (Max) 13.96% ±2.08% 14,567 $10 / $50 1050k 04 Claude Fable 5.1 (Max) 13.94% ±1.93% 17,062 $10 / $50 1M 05 GPT-6.1 Sol (Max) 13.64% ±1.63% 16,092 $2 / $10 1050k 06 Gemini 4 Argon (High) 11.75% ±0.91% 45,840 $4 / $10 1049k 07 GPT-6 Sol (Max) 8.57% ±1.71% 10,305 $2 / $10 1050k 08 Claude Opus 5 (Max) 8.22% ±1.63% 24,440 $5 / $25 1M 09 Claude Fable 5 (High) 7.75% ±1.40% 43,567 $10 / $50 1M 10 GPT-5.6 Sol (xHigh) 6.06% ±1.34% 37,542 $5 / $30 1050k 11 Claude Sonnet 5 (High) 3.75% ±1.70% 33,991 $2 / $10 1M 12 Muse Wonder (Max) 3.31% ±1.57% 11,134 $5 / $25 1049k 13 MiMo-V2.6-Pro 3.30% ±0.81% 43,983 $0.435 / $0.87 1049k 14 Gemini 3.8 Flash (High) 3.17% ±1.08% 39,719 $0.75 / $3.75 1049k 15 Muse Spark 1.3 (Max) 3.05% ±0.83% 78,017 $1.25 / $4.25 1049k 16 Grok 4.7 (xHigh) 3.02% ±1.57% 16,639 $2 / $6 500k 17 GPT-6 Luna (Max) 2.79% ±0.42% 194,437 $0.1 / $0.5 1050k 18 DeepSeek-V4.1-Flash-20260910 (Max) 2.25% ±0.34% 259,479 $0.15 / $0.6 1M 19 MiMo-V2.6-Flash 2.10% ±1.33% 22,895 $0.14 / $0.28 1049k 20 Hy4 preview 2.02% ±0.64% 93,764 $0.834 / $2.501 1024k 21 GLM-5.2 (Max) 1.69% ±0.82% 90,051 $1.4 / $4.4 1M 22 Kimi K3 (Max) 1.57% ±0.75% 139,480 $3 / $15 1049k 23 GLM-5.3 (Max) 1.22% ±0.84% 88,117 $1.4 / $4.4 1M 24 Claude Haiku 5.5 (High) 1.14% ±1.03% 27,887 $0.1 / $0.5 1M 25 Qwen3.8-Max 1.13% ±0.61% 75,307 $2 / $6 262k 26 Grok 4.6 (xHigh) 0.63% ±1.12% 29,832 $2 / $6 500k 27 GPT-5.6 Terra (xHigh) 0.22% ±1.32% 25,167 $2 / $12 1050k 28 DeepSeek-V4-Pro (High, 20260813) -0.27% ±1.73% 7,331 $1 / $2.5 1049k 29 Gemini 3.7 Flash (High) -1.27% ±0.96% 56,113 $0.75 / $3.75 1049k 30 GLM-5.3-Flash -1.77% ±0.67% 92,133 $0.15 / $0.5 1M 31 Step 5 Preview -2.03% ±1.44% 15,412 $1 / $2.7 1024k 32 Beam 1.0 501B A23B -3.04% ±2.47% 12,147 $0.25 / $1 1049k 33 Qwen3.8 Flash Next -3.27% ±0.70% 105,281 $0 / $0 262k 34 Qwen3.8 27B -4.10% ±0.79% 62,625 $0.5 / $3 1M 35 Gemini 3.1 Pro Preview -4.45% ±1.32% 87,676 $2 / $12 1049k 36 GPT-6 Luna (Low) -4.59% ±1.02% 18,454 $0.1 / $0.5 1050k 37 Mistral Large 4 -5.06% ±1.98% 14,080 $0.68 / $2.09 1049k 38 Qwen3.7-Max -6.43% ±1.16% 42,729 $2.5 / $7.5 991k 39 Hy3 -6.80% ±1.44% 31,979 $0.132 / $0.528 262k 40 Muse Spark 1.1 -6.86% ±0.65% 149,239 $1.25 / $4.25 1049k 41 Muse Spark 1.2 (xHigh) -7.02% ±0.83% 53,539 $1.25 / $4.25 1049k 42 Qwen 3.7 Plus -7.44% ±1.77% 23,985 $0.4 / $1.6 1M 43 MiMo-V2.5-Pro -8.66% ±1.00% 49,530 $1 / $3 1049k 44 MiniMax-M3 -10.07% ±1.21% 43,703 $0.3 / $1.2 1049k 45 Mistral Medium 3.5 -12.13% ±1.73% 13,468 $1.5 / $7.5 262k 46 NVIDIA-Nemotron-3.5-Super-VL-120B-A12B (NVFP4) -13.56% ±1.99% 5,172 $0.15 / $0.5 1049k 47 Nemotron 3 Ultra -14.71% ±3.94% 13,313 $0.1 / $3 262k 48 Inkling -14.90% ±1.44% 43,905 $1 / $4.05 1049k 49 Inkling Small -16.14% ±1.80% 14,421 $0.45 / $1.2 524k 50 Solar Pro 4 -20.25% ±2.16% 10,622 $0.3 / $1.2 524k 