1,516
1,496–1,537
Benchmarks / Arena (formerly LMArena)
Reported by Arena (formerly LMArena)
Arena (formerly LMArena)
Human preference on prompts that set explicit instructions.
- Results dated
- 25 Sep 2026
- Models
- 409
- Unit
- Arena rating
Full results
| # | Model | Arena Text: instruction following · 95% range Arena rating, higher is better |
|---|---|---|
| 1 | Claude Opus 5.5 (high)Anthropic | |
| 1 | Claude Opus 4.6 (high)Anthropic |
1,513 1,508–1,519 |
| 1 | Claude Fable 5 (high)Anthropic |
1,509 1,503–1,515 |
| 1 | Claude Fable 5.1 (max)Anthropic |
1,497 1,486–1,509 |
| 2 | Claude Opus 4.7 (high)Anthropic |
1,503 1,497–1,508 |
| 2 | Claude Opus 4.6Anthropic |
1,500 1,495–1,505 |
| 2 | Claude Opus 5 (high)Anthropic |
1,497 1,491–1,503 |
| 3 | Claude Opus 4.7Anthropic |
1,494 1,488–1,499 |
| 3 | Claude Opus 5 (max)Anthropic |
1,493 1,485–1,500 |
| 4 | Claude Opus 4.8 (high)Anthropic |
1,491 1,486–1,496 |
| 4 | Muse Spark 1.3 (max)Meta |
1,486 1,476–1,496 |
| 5 | GPT-5.6 Sol (xhigh)OpenAI |
1,489 1,483–1,496 |
| 5 | Gemini 3.8 Flash (high)Google |
1,488 1,480–1,496 |
| 5 | Gemini 3.7 Flash (high)Google |
1,487 1,479–1,495 |
| 6 | Kimi K3 (max)Moonshot AI |
1,486 1,479–1,493 |
| 6 | Muse Spark 1.2Meta |
1,475 1,458–1,492 |
| 7 | Claude Opus 4.5Anthropic |
1,483 1,477–1,490 |
| 7 | GLM 5.3 (max)Z.ai |
1,480 1,471–1,488 |
| 7 | DeepSeek V4.1 Flash (max)DeepSeek |
1,478 1,466–1,490 |
| 8 | MiMo-V2.6-ProXiaomi |
1,471 1,455–1,487 |
| 9 | GPT-6 Astra (max)OpenAI |
1,473 1,461–1,486 |
| 11 | Gemini 3.1 Pro PreviewGoogle |
1,479 1,475–1,484 |
| 11 | GPT-5.5 (high)OpenAI |
1,478 1,473–1,483 |
| 11 | Claude Opus 4.8Anthropic |
1,478 1,472–1,483 |
| 12 | GLM 5.3 FlashZ.ai |
1,474 1,466–1,482 |
| 12 | qwen3.7-max-previewAlibaba |
1,466 1,449–1,482 |
| 13 | GPT-5.5OpenAI |
1,475 1,470–1,480 |
| 13 | Claude Sonnet 4.6Anthropic |
1,475 1,469–1,480 |
| 13 | Claude Opus 4.5Anthropic |
1,475 1,470–1,480 |
| 13 | Muse Spark 1.1Meta |
1,473 1,467–1,479 |
| 13 | Gemini 3.6 Flash (high)Google |
1,473 1,466–1,479 |
| 13 | Qwen3.8 Max (0902)Alibaba |
1,472 1,465–1,480 |
| 15 | gemini-3-proGoogle |
1,473 1,466–1,479 |
| 15 | GPT-5.4 (high)OpenAI |
1,471 1,466–1,477 |
| 16 | GLM 5.2 (max)Z.ai |
1,470 1,464–1,476 |
| 18 | MiMo-V2.5-ProXiaomi |
1,469 1,464–1,474 |
| 18 | qwen3.5-max-previewAlibaba |
1,467 1,459–1,474 |
| 18 | Claude Sonnet 5 (high)Anthropic |
1,467 1,461–1,473 |
| 18 | muse-sparkMeta |
1,465 1,456–1,474 |
| 20 | DeepSeek V4 Pro 0813DeepSeek |
1,461 1,451–1,471 |
| 21 | Gemini 3.5 Flash (high)Google |
1,465 1,459–1,470 |
| 21 | Gemini 3.5 Flash (medium)Google |
1,464 1,458–1,470 |
| 24 | Claude Sonnet 4.5Anthropic |
1,464 1,459–1,468 |
| 24 | Grok 4.5xAI |
1,462 1,456–1,468 |
| 24 | GPT-5.6 Terra (xhigh)OpenAI |
1,462 1,455–1,468 |
| 24 | Gemma 4 31BGoogle |
1,453 1,439–1,467 |
| 28 | Claude Sonnet 4.5Anthropic |
1,462 1,457–1,466 |
| 29 | GLM 5.1Z.ai |
1,460 1,455–1,466 |
| 29 | GPT-5.4OpenAI |
1,460 1,455–1,466 |
| 31 | gemini-3-flashGoogle |
1,458 1,451–1,465 |
| 31 | MiMo-V2.6-FlashXiaomi |
1,449 1,434–1,465 |
| 32 | Claude Opus 4.1Anthropic |
1,458 1,452–1,464 |
| 33 | gpt-5.2-chat-latest-20260210OpenAI |
1,457 1,450–1,463 |
| 33 | gpt-5.5-instantOpenAI |
1,456 1,449–1,463 |
| 33 | Kimi K2.6Moonshot AI |
1,456 1,449–1,462 |
| 33 | Qwen3.6 Max PreviewAlibaba |
1,449 1,435–1,463 |
| 33 | GPT-6 Sol (max)OpenAI |
1,447 1,432–1,461 |
| 37 | Claude Opus 4.1Anthropic |
1,454 1,450–1,459 |
| 38 | ernie-5.1Baidu |
1,452 1,445–1,459 |
| 38 | Grok 4.6 (high)xAI |
1,451 1,443–1,458 |
| 40 | DeepSeek V4 Pro 0423DeepSeek |
1,452 1,447–1,458 |
| 40 | GPT-6 Luna (max)OpenAI |
1,443 1,429–1,457 |
| 41 | GPT-5.1 (high)OpenAI |
1,450 1,444–1,457 |
| 41 | grok-4.20-beta1xAI |
1,450 1,443–1,457 |
| 42 | Hy3Tencent |
1,446 1,435–1,456 |
| 47 | deepseek-v4-pro-preview (high reasoning)DeepSeek |
1,448 1,442–1,454 |
| 47 | Qwen3.7 PlusAlibaba |
1,447 1,441–1,453 |
| 47 | GLM 5Z.ai |
1,446 1,439–1,452 |
| 47 | Grok 4.7 (xhigh)xAI |
1,437 1,421–1,453 |
| 48 | grok-4.20-beta-0309-reasoningxAI |
1,446 1,441–1,452 |
| 48 | GPT-5.6 Luna (xhigh)OpenAI |
1,446 1,439–1,452 |
| 48 | mimo-v2-proXiaomi |
1,445 1,437–1,452 |
| 48 | claude-opus-4 (thinking-16k reasoning)Anthropic |
1,444 1,437–1,451 |
| 48 | Gemma 4 26B A4BGoogle |
1,437 1,424–1,451 |
| 51 | Gemini 3.5 Flash LiteGoogle |
1,444 1,437–1,450 |
| 54 | grok-4.20-multi-agent-beta-0309xAI |
1,444 1,438–1,449 |
| 55 | gemini-3-flash (minimal reasoning)Google |
1,443 1,438–1,448 |
| 57 | Kimi K2.5Moonshot AI |
1,439 1,434–1,444 |
| 57 | gpt-4.5-preview-2025-02-27OpenAI |
1,437 1,428–1,445 |
| 58 | Kimi K2.5Moonshot AI |
1,432 1,420–1,444 |
| 60 | gpt-5.3-chat-latestOpenAI |
1,436 1,429–1,442 |
| 61 | Gemini 2.5 ProGoogle |
1,438 1,434–1,442 |
| 61 | Qwen3.6 PlusAlibaba |
1,436 1,430–1,442 |
| 63 | dola-seed-2.0-proBytedance |
1,435 1,430–1,440 |
| 63 | MiniMax M3MiniMax |
1,435 1,429–1,440 |
| 63 | DeepSeek V3.1 TerminusDeepSeek |
1,420 1,400–1,440 |
| 65 | GPT-5.4 Mini (high)OpenAI |
1,434 1,428–1,439 |
| 66 | Qwen3.5 397B A17BAlibaba |
1,434 1,429–1,439 |
| 66 | deepseek-v4-flash-preview (high reasoning)DeepSeek |
1,433 1,428–1,439 |
| 71 | grok-4.1xAI |
1,431 1,426–1,436 |
| 71 | MiMo-V2.5Xiaomi |
1,430 1,424–1,437 |
| 71 | GLM 4.7Z.ai |
1,425 1,415–1,436 |
| 73 | grok-4.1-thinkingxAI |
1,429 1,424–1,434 |
| 73 | mimo-v2-omniXiaomi |
1,427 1,419–1,435 |
| 75 | DeepSeek V4 Flash 0423DeepSeek |
1,428 1,422–1,434 |
| 76 | GPT-5.2 (high)OpenAI |
1,427 1,421–1,433 |
| 76 | InklingThinkingmachines |
1,427 1,420–1,433 |
| 76 | GPT-5.1OpenAI |
1,426 1,420–1,432 |
| 76 | GLM 5V TurboZ.ai |
1,423 1,413–1,433 |
| 76 | amazon-nova-experimental-chat-26-02-10Amazon |
1,414 1,397–1,432 |
| 77 | chatgpt-4o-latest-20250326OpenAI |
1,427 1,422–1,431 |
| 77 | ernie-5.0-0110Baidu |
1,425 1,419–1,431 |
| 77 | qwen3-max-previewAlibaba |
1,424 1,417–1,432 |
| 77 | Qwen3.8 27BAlibaba |
1,423 1,415–1,432 |
| 77 | Mistral Medium 3.5Mistral |
1,421 1,411–1,431 |
| 77 | ernie-5.0-preview-1203Baidu |
1,419 1,408–1,431 |
| 83 | GPT-5.2OpenAI |
1,424 1,419–1,429 |
| 85 | muse-glimmerMeta |
1,412 1,396–1,428 |
| 86 | DeepSeek V3.2DeepSeek |
1,420 1,414–1,426 |
| 86 | DeepSeek V3.1DeepSeek |
1,416 1,405–1,427 |
| 86 | DeepSeek V3.2 ExpDeepSeek |
1,416 1,404–1,427 |
| 87 | DeepSeek V3.2DeepSeek |
1,420 1,414–1,425 |
| 87 | DeepSeek V3.2 ExpDeepSeek |
1,414 1,404–1,424 |
| 87 | Qwen3 MaxAlibaba |
1,413 1,402–1,425 |
| 90 | kimi-k2-thinking-turboMoonshot AI |
1,418 1,413–1,423 |
| 90 | longcat-flash-chat-2602-expMeituan |
1,416 1,409–1,423 |
| 92 | Claude Sonnet 4Anthropic |
1,415 1,408–1,422 |
| 92 | claude-opus-4Anthropic |
1,415 1,409–1,422 |
| 92 | gpt-5-chatOpenAI |
1,415 1,408–1,422 |
| 94 | Qwen3 VL 235B A22B InstructAlibaba |
1,409 1,398–1,421 |
| 96 | Grok 4.3xAI |
1,415 1,410–1,420 |
| 99 | Claude Haiku 4.5Anthropic |
1,415 1,411–1,419 |
| 100 | Qwen3 235B A22B Instruct 2507Alibaba |
1,414 1,410–1,419 |
| 100 | GLM 4.6Z.ai |
1,412 1,406–1,419 |
| 100 | claude-3-7-sonnet-20250219 (thinking-32k reasoning)Anthropic |
1,412 1,405–1,418 |
| 100 | GPT-5 (high)OpenAI |
1,410 1,403–1,417 |
| 100 | hunyuan-vision-1.5-thinkingTencent |
1,396 1,374–1,419 |
| 101 | amazon-nova-experimental-chat-26-01-10Amazon |
1,398 1,380–1,416 |
| 103 | nvidia-nemotron-3-ultra-550b-a55b-nvfp4NVIDIA |
1,404 1,394–1,414 |
| 104 | o1OpenAI |
1,407 1,401–1,414 |
| 107 | Gemini 3.1 Flash Lite PreviewGoogle |
1,407 1,402–1,413 |
| 107 | MiniMax M2.7MiniMax |
1,407 1,402–1,412 |
| 107 | Qwen3.5-122B-A10BAlibaba |
1,406 1,399–1,412 |
| 107 | GLM 4.5Z.ai |
1,405 1,397–1,413 |
| 107 | DeepSeek V3.1DeepSeek |
1,402 1,392–1,412 |
| 107 | ernie-5.0-preview-1022Baidu |
1,398 1,382–1,413 |
| 107 | DeepSeek V3.1 TerminusDeepSeek |
1,394 1,377–1,412 |
| 110 | grok-3-preview-02-24xAI |
1,403 1,397–1,410 |
| 110 | grok-4-fast-chatxAI |
1,395 1,381–1,410 |
| 111 | GPT-4.1OpenAI |
1,403 1,398–1,409 |
| 111 | Hy3 previewTencent |
1,397 1,385–1,410 |
| 115 | o3OpenAI |
1,403 1,397–1,408 |
| 116 | grok-4-1-fast-reasoningxAI |
1,401 1,396–1,407 |
| 116 | Qwen3.5-27BAlibaba |
1,400 1,394–1,407 |
| 116 | gemini-2.5-flash-preview-09-2025Google |
1,400 1,393–1,406 |
| 119 | Mistral Large 3 2512Mistral |
1,400 1,396–1,405 |
| 119 | R1DeepSeek |
1,397 1,390–1,405 |
| 121 | Gemini 2.5 FlashGoogle |
1,400 1,396–1,403 |
| 121 | grok-4-0709xAI |
1,397 1,391–1,403 |
| 122 | Claude Sonnet 4Anthropic |
1,396 1,389–1,402 |
| 125 | Kimi K2 0905Moonshot AI |
1,390 1,379–1,401 |
| 126 | Inkling SmallThinkingmachines |
1,393 1,386–1,400 |
| 126 | R1 0528DeepSeek |
1,391 1,381–1,400 |
| 127 | Mistral Medium 3.1Mistral |
1,395 1,391–1,399 |
| 128 | longcat-flash-chatMeituan |
1,388 1,377–1,399 |
| 130 | Qwen3 235B A22B Thinking 2507Alibaba |
1,385 1,372–1,397 |
| 131 | grok-4-fast-reasoningxAI |
1,389 1,380–1,397 |
| 136 | amazon-nova-experimental-chat-12-10Amazon |
1,377 1,359–1,396 |
| 138 | Qwen3 VL 235B A22B ThinkingAlibaba |
1,383 1,370–1,395 |
| 140 | minimax-m2.1-previewMiniMax |
1,385 1,376–1,394 |
| 141 | Qwen3.5-35B-A3BAlibaba |
1,386 1,380–1,393 |
| 142 | GPT-5.4 Nano (high)OpenAI |
1,387 1,381–1,392 |
| 142 | claude-3-7-sonnet-20250219Anthropic |
1,385 1,380–1,391 |
| 142 | Qwen3 Coder 480B A35BAlibaba |
1,384 1,376–1,392 |
| 142 | hunyuan-t1-20250711Tencent |
1,374 1,356–1,391 |
| 145 | Step 3.5 FlashStepfun |
1,384 1,379–1,389 |
| 146 | mimo-v2-flash (no reasoning)Xiaomi |
1,382 1,376–1,388 |
| 146 | Qwen3.5-FlashAlibaba |
1,381 1,376–1,387 |
| 146 | MiniMax M2.5MiniMax |
1,381 1,375–1,387 |
| 146 | o1-previewOpenAI |
1,381 1,374–1,388 |
| 146 | Kimi K2 0711Moonshot AI |
1,381 1,373–1,389 |
| 146 | Qwen3 235B A22BAlibaba |
1,381 1,374–1,388 |
| 146 | Solar Pro 4Upstage |
1,379 1,370–1,388 |
| 146 | mimo-v2-flash (thinking reasoning)Xiaomi |
1,375 1,365–1,386 |
| 146 | GLM 4.6VZ.ai |
1,367 1,346–1,388 |
| 148 | DeepSeek V3 0324DeepSeek |
1,378 1,372–1,384 |
| 148 | Qwen3 Next 80B A3B InstructAlibaba |
1,376 1,369–1,384 |
| 152 | GPT-5 Mini (high)OpenAI |
1,373 1,365–1,381 |
| 156 | GPT-4.1 MiniOpenAI |
1,373 1,366–1,379 |
| 158 | claude-3-5-sonnet-20241022Anthropic |
1,374 1,370–1,378 |
| 161 | trinity-large-previewArcee Ai |
1,370 1,363–1,376 |
| 165 | o4 MiniOpenAI |
1,369 1,362–1,375 |
| 165 | gemini-2.5-flash-lite-preview-06-17 (thinking reasoning)Google |
1,368 1,361–1,375 |
| 165 | Qwen3 30B A3B Instruct 2507Alibaba |
1,366 1,359–1,374 |
| 167 | o3 Mini HighOpenAI |
1,366 1,359–1,374 |
| 168 | Mistral Medium 3Mistral |
1,366 1,359–1,373 |
| 169 | llama-3.1-nemotron-ultra-253b-v1NVIDIA |
1,351 1,330–1,373 |
| 174 | gemini-2.5-flash-lite-preview-09-2025 (no reasoning)Google |
1,364 1,358–1,369 |
| 174 | grok-3-mini (high reasoning)xAI |
1,360 1,351–1,369 |
| 174 | Qwen3 Next 80B A3B ThinkingAlibaba |
1,359 1,349–1,369 |
| 175 | GLM 4.5 AirZ.ai |
1,361 1,354–1,368 |
| 175 | amazon-nova-experimental-chat-11-10Amazon |
1,361 1,354–1,368 |
| 175 | hunyuan-turbos-20250226Tencent |
1,351 1,333–1,368 |
| 177 | hunyuan-turbos-20250416Tencent |
1,353 1,341–1,365 |
| 178 | Trinity Large ThinkingArcee Ai |
1,357 1,351–1,364 |
| 178 | Qwen3 235B A22BAlibaba |
1,357 1,349–1,365 |
| 179 | qwen2.5-maxAlibaba |
1,357 1,351–1,363 |
| 181 | grok-3-mini-betaxAI |
1,353 1,344–1,361 |
| 183 | step-3Stepfun |
1,344 1,330–1,359 |
| 185 | GLM 4.7 FlashZ.ai |
1,348 1,338–1,358 |
| 186 | granite-4.2-30bIBM |
1,339 1,321–1,357 |
| 187 | amazon-nova-experimental-chat-10-20Amazon |
1,345 1,335–1,356 |
| 187 | Nemotron 3 SuperNVIDIA |
1,343 1,330–1,355 |
| 187 | GLM 4.5VZ.ai |
1,338 1,323–1,354 |
| 189 | gemini-2.0-flash-001Google |
1,347 1,342–1,353 |
| 189 | MiniMax M1MiniMax |
1,346 1,340–1,353 |
| 189 | nvidia-nemotron-3.5-lightning-30b-a3b-nvfp4NVIDIA |
1,343 1,333–1,353 |
| 192 | DeepSeek V3DeepSeek |
1,343 1,336–1,350 |
| 192 | Qwen3 32BAlibaba |
1,331 1,312–1,350 |
| 194 | o3 MiniOpenAI |
1,344 1,339–1,349 |
| 194 | Gemma 3 27BGoogle |
1,343 1,337–1,349 |
| 194 | Command ACohere |
1,342 1,336–1,347 |
| 194 | Mistral Small 3.2 24BMistral |
1,338 1,329–1,347 |
| 194 | MiniMax M2MiniMax |
1,334 1,321–1,347 |
| 194 | intellect-3Primeintellect |
1,332 1,317–1,348 |
| 194 | Mercury 2Inception |
1,327 1,307–1,346 |
| 195 | gemini-1.5-pro-002Google |
1,340 1,335–1,345 |
| 195 | llama-3.3-nemotron-49b-super-v1NVIDIA |
1,325 1,306–1,345 |
| 196 | claude-3-5-sonnet-20240620Anthropic |
1,338 1,333–1,343 |
| 196 | Qwen-PlusAlibaba |
1,332 1,320–1,344 |
| 196 | molmo-2-8bAi2 |
1,304 1,267–1,342 |
| 198 | GPT-5 Nano (high)OpenAI |
1,327 1,314–1,340 |
| 198 | nvidia-llama-3.3-nemotron-super-49b-v1.5NVIDIA |
1,321 1,301–1,341 |
| 201 | gemini-2.0-flash-lite-preview-02-05Google |
1,331 1,325–1,338 |
| 201 | Nova 2 LiteAmazon |
1,327 1,317–1,338 |
| 202 | o1-miniOpenAI |
1,332 1,326–1,337 |
| 202 | Gemma 3 12BGoogle |
1,321 1,305–1,337 |
| 204 | amazon-nova-experimental-chat-10-09Amazon |
1,315 1,293–1,336 |
| 209 | GPT-4o (2024-05-13)OpenAI |
1,326 1,321–1,331 |
| 209 | gpt-oss-120bOpenAI |
1,323 1,316–1,330 |
| 209 | olmo-3.1-32b-instructAi2 |
1,320 1,310–1,331 |
| 209 | hunyuan-large-2025-02-10Tencent |
1,316 1,300–1,332 |
| 209 | hunyuan-turbo-0110Tencent |
1,314 1,295–1,332 |
| 210 | ring-flash-2.0Ant Group |
1,316 1,303–1,330 |
| 211 | qwq-32bAlibaba |
1,323 1,316–1,330 |
| 212 | ling-flash-2.0Ant Group |
1,315 1,302–1,329 |
| 213 | step-2-16k-exp-202412Stepfun |
1,316 1,303–1,328 |
| 213 | glm-4-plus-0111Z.ai |
1,315 1,302–1,327 |
| 214 | deepseek-v2.5-1210DeepSeek |
1,314 1,303–1,326 |
| 215 | GPT-4o (2024-08-06)OpenAI |
1,319 1,313–1,325 |
| 215 | gemini-advanced-0514Google |
1,317 1,311–1,324 |
| 216 | claude-3-5-haiku-20241022Anthropic |
1,316 1,311–1,321 |
| 219 | llama-3.1-405b-instruct-fp8Meta |
1,315 1,310–1,320 |
| 219 | llama-3.1-405b-instruct-bf16Meta |
1,314 1,309–1,320 |
| 219 | Llama 4 MaverickMeta |
1,314 1,307–1,320 |
| 219 | step-1o-turbo-202506Stepfun |
1,307 1,294–1,320 |
| 220 | claude-3-opus-20240229Anthropic |
1,314 1,310–1,319 |
| 220 | grok-2-2024-08-13xAI |
1,312 1,307–1,317 |
| 220 | gemini-1.5-pro-001Google |
1,312 1,306–1,317 |
| 220 | Qwen3 30B A3BAlibaba |
1,310 1,302–1,318 |
| 224 | yi-lightning01 Ai |
1,308 1,301–1,315 |
| 224 | olmo-3-32b-thinkAi2 |
1,298 1,283–1,314 |
| 225 | GPT-4.1 NanoOpenAI |
1,300 1,288–1,313 |
| 228 | magistral-medium-2506Mistral |
1,300 1,290–1,311 |
| 232 | qwen-max-0919Alibaba |
1,302 1,294–1,309 |
| 233 | GPT-4 TurboOpenAI |
1,303 1,298–1,309 |
| 233 | athene-v2-chatNexusflow |
1,302 1,296–1,308 |
| 233 | glm-4-plusZ.ai |
1,301 1,294–1,308 |
| 236 | Llama 4 ScoutMeta |
1,299 1,292–1,306 |
| 238 | Mistral Large 2407Mistral |
1,299 1,294–1,305 |
| 239 | qwen2.5-plus-1127Alibaba |
1,294 1,285–1,304 |
| 239 | granite-4.1-8bIBM |
1,287 1,270–1,305 |
| 239 | hunyuan-large-visionTencent |
1,287 1,270–1,303 |
| 239 | granite-4.2-3bIBM |
1,283 1,264–1,303 |
| 242 | gpt-4-1106-previewOpenAI |
1,297 1,291–1,302 |
| 247 | mistral-large-2411Mistral |
1,295 1,289–1,301 |
| 247 | Mistral Small 3.1 24BMistral |
1,294 1,287–1,301 |
| 248 | GPT-4o-mini (2024-07-18)OpenAI |
1,293 1,288–1,298 |
| 248 | deepseek-v2.5DeepSeek |
1,292 1,285–1,298 |
| 248 | Nemotron 3 Nano 30B A3BNVIDIA |
1,291 1,282–1,300 |
| 248 | Granite 4.2 8BIBM |
1,280 1,261–1,299 |
| 249 | Llama 3.3 70B InstructMeta |
1,293 1,288–1,298 |
| 249 | Qwen2.5 72B InstructAlibaba |
1,292 1,286–1,298 |
| 249 | gpt-4-0125-previewOpenAI |
1,290 1,285–1,296 |
| 249 | gemini-1.5-flash-002Google |
1,290 1,284–1,296 |
| 249 | hunyuan-standard-2025-02-10Tencent |
1,282 1,266–1,297 |
| 251 | mercuryInception |
1,269 1,243–1,294 |
| 254 | gpt-oss-20bOpenAI |
1,281 1,269–1,294 |
| 256 | gpt-4-0314OpenAI |
1,285 1,278–1,293 |
| 256 | olmo-3.1-32b-thinkAi2 |
1,279 1,266–1,293 |
| 257 | llama-3.1-nemotron-70b-instructNVIDIA |
1,280 1,269–1,291 |
| 258 | gemma-3n-e4b-itGoogle |
1,282 1,273–1,290 |
| 259 | grok-2-mini-2024-08-13xAI |
1,284 1,278–1,289 |
| 262 | llama-3.1-tulu-3-70bAi2 |
1,272 1,257–1,288 |
| 264 | athene-70b-0725Nexusflow |
1,279 1,271–1,286 |
| 269 | gpt-4-0613OpenAI |
1,277 1,271–1,283 |
| 269 | Nova Pro 1.0Amazon |
1,277 1,270–1,283 |
| 269 | ibm-granite-h-smallIBM |
1,268 1,252–1,283 |
| 269 | Gemma 3 4BGoogle |
1,267 1,251–1,284 |
| 272 | Llama 3.1 70B InstructMeta |
1,273 1,267–1,278 |
| 273 | Gemma 2 27BGoogle |
1,271 1,267–1,276 |
| 273 | claude-3-sonnet-20240229Anthropic |
1,268 1,263–1,274 |
| 273 | jamba-1.5-largeAI21 Labs |
1,267 1,256–1,277 |
| 273 | Qwen2.5 Coder 32B InstructAlibaba |
1,264 1,252–1,276 |
| 273 | llama-3.1-nemotron-51b-instructNVIDIA |
1,263 1,248–1,277 |
| 274 | gemini-1.5-flash-001Google |
1,265 1,259–1,271 |
| 274 | reka-core-20240904Rekaai |
1,262 1,252–1,272 |
| 281 | nemotron-4-340b-instructNVIDIA |
1,259 1,251–1,268 |
| 282 | glm-4-0520Z.ai |
1,256 1,246–1,267 |
| 286 | llama-3-70b-instructMeta |
1,259 1,253–1,264 |
| 286 | Mistral Small 3Mistral |
1,255 1,247–1,264 |
| 287 | Command R+ (08-2024)Cohere |
1,254 1,245–1,263 |
| 288 | deepseek-coder-v2DeepSeek |
1,253 1,244–1,262 |
| 288 | gemma-2-9b-it-simpoPrinceton Nlp |
1,252 1,242–1,262 |
| 289 | reka-flash-20240904Rekaai |
1,249 1,239–1,260 |
| 289 | hunyuan-standard-256kTencent |
1,243 1,227–1,260 |
| 292 | c4ai-aya-expanse-32bCohere |
1,249 1,243–1,256 |
| 295 | Phi 4Microsoft |
1,245 1,239–1,252 |
| 298 | claude-3-haiku-20240307Anthropic |
1,245 1,240–1,251 |
| 298 | gemma-2-9b-itGoogle |
1,245 1,240–1,251 |
| 298 | Nova Lite 1.0Amazon |
1,244 1,237–1,251 |
| 298 | qwen2-72b-instructAlibaba |
1,243 1,236–1,249 |
| 299 | command-r-plusCohere |
1,241 1,235–1,247 |
| 300 | olmo-2-0325-32b-instructAi2 |
1,229 1,212–1,246 |
| 301 | Command R (08-2024)Cohere |
1,236 1,227–1,245 |
| 302 | gemini-1.5-flash-8b-001Google |
1,238 1,232–1,244 |
| 302 | mistral-large-2402Mistral |
1,238 1,231–1,244 |
| 310 | gemini-proGoogle |
1,220 1,203–1,237 |
| 317 | qwen1.5-110b-chatAlibaba |
1,218 1,210–1,226 |
| 317 | gpt-3.5-turbo-0125OpenAI |
1,216 1,209–1,222 |
| 317 | Mixtral 8x22B InstructMistral |
1,215 1,209–1,222 |
| 317 | Nova Micro 1.0Amazon |
1,215 1,208–1,222 |
| 317 | qwen1.5-72b-chatAlibaba |
1,212 1,205–1,219 |
| 317 | ministral-8b-2410Mistral |
1,211 1,198–1,224 |
| 317 | mistral-mediumMistral |
1,211 1,202–1,219 |
| 317 | gemini-pro-dev-apiGoogle |
1,210 1,200–1,221 |
| 317 | llama-3.1-tulu-3-8bAi2 |
1,208 1,193–1,224 |
| 317 | jamba-1.5-miniAI21 Labs |
1,205 1,194–1,215 |
| 317 | c4ai-aya-expanse-8bCohere |
1,205 1,195–1,214 |
| 317 | reka-flash-21b-20240226-onlineRekaai |
1,202 1,192–1,212 |
| 319 | gpt-3.5-turbo-1106OpenAI |
1,198 1,186–1,210 |
| 319 | granite-3.1-8b-instructIBM |
1,193 1,176–1,210 |
| 320 | zephyr-orpo-141b-A35b-v0.1Huggingface |
1,194 1,178–1,209 |
| 322 | command-rCohere |
1,200 1,193–1,207 |
| 325 | reka-flash-21b-20240226Rekaai |
1,193 1,184–1,201 |
| 326 | llama-3-8b-instructMeta |
1,193 1,187–1,198 |
| 327 | Llama 3.1 8B InstructMeta |
1,191 1,185–1,197 |
| 327 | yi-1.5-34b-chat01 Ai |
1,188 1,180–1,196 |
| 327 | dbrx-instruct-previewDatabricks |
1,188 1,179–1,196 |
| 329 | qwen1.5-32b-chatAlibaba |
1,185 1,176–1,193 |
| 332 | mixtral-8x7b-instruct-v0.1Mistral |
1,181 1,174–1,187 |
| 332 | internlm2_5-20b-chatShanghai Ai Lab |
1,178 1,168–1,188 |
| 332 | granite-3.1-2b-instructIBM |
1,173 1,157–1,190 |
| 334 | tulu-2-dpo-70bAi2 |
1,171 1,156–1,186 |
| 335 | phi-3-medium-4k-instructMicrosoft |
1,178 1,170–1,185 |
| 335 | granite-3.0-8b-instructIBM |
1,172 1,160–1,185 |
| 338 | qwen1.5-14b-chatAlibaba |
1,168 1,158–1,178 |
| 339 | gemma-2-2b-itGoogle |
1,171 1,165–1,177 |
| 339 | wizardlm-70bMicrosoft |
1,164 1,150–1,177 |
| 341 | deepseek-llm-67b-chatDeepSeek |
1,158 1,141–1,175 |
| 343 | openchat-3.5Openchat |
1,155 1,140–1,169 |
| 343 | openhermes-2.5-mistral-7bTeknium |
1,154 1,138–1,169 |
| 344 | openchat-3.5-0106Openchat |
1,156 1,145–1,166 |
| 345 | gemma-1.1-7b-itGoogle |
1,157 1,149–1,165 |
| 345 | phi-3-small-8k-instructMicrosoft |
1,154 1,146–1,163 |
| 345 | snowflake-arctic-instructSnowflake |
1,154 1,145–1,162 |
| 345 | yi-34b-chat01 Ai |
1,152 1,143–1,162 |
| 345 | qwq-32b-previewAlibaba |
1,146 1,130–1,161 |
| 345 | falcon-180b-chatTII |
1,135 1,106–1,163 |
| 347 | Llama 3.2 3B InstructMeta |
1,146 1,135–1,157 |
| 349 | starling-lm-7b-betaNexusflow |
1,146 1,136–1,156 |
| 349 | mpt-30b-chatMosaicml |
1,135 1,114–1,156 |
| 349 | dolphin-2.2.1-mistral-7bCognitivecomputations |
1,129 1,105–1,154 |
| 350 | vicuna-33bLmsys |
1,140 1,131–1,149 |
| 351 | starling-lm-7b-alphaBerkeley |
1,137 1,125–1,148 |
| 354 | llama-2-70b-chatMeta |
1,135 1,127–1,143 |
| 354 | granite-3.0-2b-instructIBM |
1,132 1,120–1,145 |
| 354 | llama2-70b-steerlm-chatNVIDIA |
1,125 1,106–1,143 |
| 357 | wizardlm-13bMicrosoft |
1,124 1,110–1,138 |
| 357 | qwen1.5-7b-chatAlibaba |
1,124 1,110–1,138 |
| 359 | qwen-14b-chatAlibaba |
1,119 1,103–1,135 |
| 360 | phi-3-mini-4k-instruct-june-2024Microsoft |
1,125 1,115–1,134 |
| 360 | mistral-7b-instruct-v0.2Mistral |
1,123 1,114–1,132 |
| 360 | solar-10.7b-instruct-v1.0Upstage |
1,116 1,098–1,135 |
| 361 | palm-2Google |
1,117 1,104–1,131 |
| 362 | codellama-70b-instructMeta |
1,098 1,068–1,127 |
| 363 | vicuna-13bLmsys |
1,117 1,108–1,127 |
| 363 | smollm2-1.7b-instructHuggingface |
1,104 1,083–1,125 |
| 364 | phi-3-mini-4k-instructMicrosoft |
1,113 1,105–1,122 |
| 364 | nous-hermes-2-mixtral-8x7b-dpoNousresearch |
1,108 1,091–1,124 |
| 364 | zephyr-7b-alphaHuggingface |
1,099 1,075–1,123 |
| 365 | llama-2-13b-chatMeta |
1,110 1,101–1,120 |
| 365 | gemma-7b-itGoogle |
1,105 1,092–1,118 |
| 365 | codellama-34b-instructMeta |
1,104 1,091–1,117 |
| 368 | gemma-1.1-2b-itGoogle |
1,101 1,091–1,112 |
| 368 | phi-3-mini-128k-instructMicrosoft |
1,100 1,090–1,111 |
| 372 | stripedhyena-nous-7bTogether |
1,091 1,076–1,106 |
| 377 | zephyr-7b-betaHuggingface |
1,089 1,076–1,102 |
| 378 | mistral-7b-instructMistral |
1,087 1,074–1,101 |
| 379 | Llama 3.2 1B InstructMeta |
1,086 1,074–1,097 |
| 382 | vicuna-7bLmsys |
1,077 1,063–1,091 |
| 384 | gemma-2b-itGoogle |
1,071 1,055–1,087 |
| 384 | qwen1.5-4b-chatAlibaba |
1,070 1,057–1,083 |
| 384 | guanaco-33bTimdettmers |
1,066 1,045–1,087 |
| 385 | llama-2-7b-chatMeta |
1,070 1,060–1,080 |
| 391 | gpt4all-13b-snoozyNomic |
1,038 1,013–1,063 |
| 395 | chatglm3-6bZ.ai |
1,037 1,019–1,054 |
| 395 | olmo-7b-instructAi2 |
1,029 1,013–1,045 |
| 396 | koala-13bBerkeley |
1,027 1,011–1,042 |
| 396 | alpaca-13bStanford |
1,025 1,009–1,041 |
| 396 | mpt-7b-chatMosaicml |
1,009 991–1,028 |
| 401 | oasst-pythia-12bOpenassistant |
988 972–1,004 |
| 401 | chatglm2-6bZ.ai |
983 960–1,005 |
| 401 | chatglm-6bZ.ai |
978 960–996 |
| 402 | RWKV-4-Raven-14BRwkv |
970 953–987 |
| 402 | fastchat-t5-3bLmsys |
958 940–977 |
| 403 | dolly-v2-12bDatabricks |
940 918–961 |
| 406 | llama-13bMeta |
918 893–944 |
| 407 | stablelm-tuned-alpha-7bStabilityai |
910 890–930 |
Models share a rank when their ranges overlap. Results as published by Arena (formerly LMArena); we do not re-run them.
What it measures
Human preference on prompts that set explicit instructions.
What it does not measure
Not accuracy or correctness: it ranks which answer voters preferred, with answer length and formatting controlled for.
Source
Contains data from the Arena Leaderboard Dataset by Arena, licensed under CC BY 4.0. Licence: Creative Commons Attribution 4.0 International.