Leaderboards

Benchmarks

Every score carries its origin — leaderboard is independent,vendor is self-reported. History is kept, never overwritten.

Reasoning280Coding266Math174Agentic110Chat515Speech654
Arena Creative Writing143Arena Instruction Following143Arena Text (overall)143IFEval16LiveBench Instruction Following35LiveBench Language35

Arena Instruction Following

Arena Score on instruction-following prompts — not a percentagehigher is better
Provenanceindependent leaderboard100 modelsscores Aug 2, 2026
1Claude Fable 5Anthropic1515.023230Aug 2, 2026leaderboard2Claude Opus 4.6Anthropic1499.39760Aug 2, 2026leaderboard3Claude Opus 5Anthropic1497.8623via High60Aug 2, 2026leaderboard4Claude Opus 4.7Anthropic1490.936360Aug 2, 2026leaderboard5GPT-5.6 SolOpenAI1481.0269via xHigh99Aug 2, 2026leaderboard6Gemini 3.1 Pro Preview Custom ToolsGoogle1480.5518247Aug 2, 2026leaderboard7Kimi K3Moonshot AI2.8T1479.7004via Max106~0.84Aug 2, 2026leaderboard8Claude Opus 4.8Anthropic1477.27259Aug 2, 2026leaderboard9Claude Sonnet 4.6Anthropic1477.102998Aug 2, 2026leaderboard10Gemini 3.6 FlashGoogle1476.7944394Aug 2, 2026leaderboard11Muse Spark 1.1Meta1475.302347Aug 2, 2026leaderboard12Claude Opus 4.5Anthropic1474.312759Aug 2, 2026leaderboard13GPT-5.5OpenAI1472.7279147Aug 2, 2026leaderboard14MiMo-V2.5-ProXiaomi1T1468.5762110~2.3Aug 2, 2026leaderboard15Qwen3.7 MaxQwen1466.4005391Aug 2, 2026leaderboard16GLM 5.1Z.AI754B1464.2909489~3.1Aug 2, 2026leaderboard17Gemini 3.5 FlashGoogle1464.0948via High325Aug 2, 2026leaderboard18GLM 5.2Z.AI753B1464.0067via Max976~3.1Aug 2, 2026leaderboard19GPT-5.6 TerraOpenAI1463.7747via xHigh488Aug 2, 2026leaderboard20Grok 4.5xAI1463.1179244Aug 2, 2026leaderboard21Claude Sonnet 4.5Anthropic1462.634998Aug 2, 2026leaderboard22GPT-5.4OpenAI1461.6139195Aug 2, 2026leaderboard23GPT-5.2 ChatOpenAI1457.0307208Aug 2, 2026leaderboard24Claude Opus 4.1Anthropic1455.004219Aug 2, 2026leaderboard25DeepSeek V4 ProDeepSeek1.6T1454.67741672~1.4Aug 2, 2026leaderboard26Kimi K2.6Moonshot AI1.1T1454.1993586~2.2Aug 2, 2026leaderboard27Gemma 4 31BGoogle32.7B1451.62114269~70Aug 2, 2026leaderboard28Qwen3.7 PlusQwen1451.01131134Aug 2, 2026leaderboard29Qwen3.6 Max PreviewQwen1449.2104235Aug 2, 2026leaderboard30Grok 4.20xAI1448.9409580Aug 2, 2026leaderboard31GPT-5.6 LunaOpenAI1448.0055via xHigh1097Aug 2, 2026leaderboard32Hy3Tencent299B1446.98042740~7.7Aug 2, 2026leaderboard33GLM 5Z.AI754B1446.644753~3Aug 2, 2026leaderboard34Grok 4.20 Multi-AgentxAI1444.5373578Aug 2, 2026leaderboard35Gemma 4 26B A4BGoogle26.5B1440.55224802~86Aug 2, 2026leaderboard36MiniMax M3MiniMax427B1439.37611499~5.3Aug 2, 2026leaderboard37Gemini 2.5 ProGoogle1439.233288Aug 2, 2026leaderboard38Gemini 3.5 Flash LiteGoogle1438.91021151Aug 2, 2026leaderboard39Kimi K2 ThinkingMoonshot AI1.1T1438.8109via Thinking576~2.2Aug 2, 2026leaderboard40Qwen3.6 PlusQwen1436.4654737Aug 2, 2026leaderboard41GPT-5.4 MiniOpenAI1435.066via High638Aug 2, 2026leaderboard42Kimi K2.5Moonshot AI1.1T1434.4493717~2.1Aug 2, 2026leaderboard43GPT-5.3 ChatOpenAI1434.2717102Aug 2, 2026leaderboard44MiMo-V2.5Xiaomi311B1431.33096390~7.3Aug 2, 2026leaderboard45Qwen3.5 397B A17BQwen403B1431.2875612~5.6Aug 2, 2026leaderboard46DeepSeek V4 FlashDeepSeek291B1428.27948097~7.8Aug 2, 2026leaderboard47GLM 4.7Z.AI358B1427.809816~6.3Aug 2, 2026leaderboard48GPT-5.1OpenAI1427.0893285Aug 2, 2026leaderboard49Qwen3 MaxQwen1426.6421366Aug 2, 2026leaderboard50GLM 5V TurboZ.AI1422.2167356Aug 2, 2026leaderboard51Mistral Medium 3.5Mistral AI128B1421.7924190~18Aug 2, 2026leaderboard52DeepSeek V3.2DeepSeek685B1420.55574571~3.3Aug 2, 2026leaderboard53InklingThinking Machines952B1420.3974351~2.4Aug 2, 2026leaderboard54DeepSeek V3.2 ExpDeepSeek685B1416.1123454~3.3Aug 2, 2026leaderboard55Qwen3 235B A22B Instruct 2507Qwen235B1415.86462574~9.6Aug 2, 2026leaderboard56Claude Haiku 4.5Anthropic1415.3539283Aug 2, 2026leaderboard57GLM 4.6Z.AI357B1414.9819809~6.3Aug 2, 2026leaderboard58Grok 4.3xAI1414.4779566Aug 2, 2026leaderboard59Claude Opus 4Anthropic1413.753419Aug 2, 2026leaderboard60Qwen3 VL 235B A22B InstructQwen236B1413.28171606~9.5Aug 2, 2026leaderboard61MiniMax M2.7MiniMax229B1409.71121468~9.8Aug 2, 2026leaderboard62o1OpenAI1407.075623Aug 2, 2026leaderboard63Qwen3.5-122B-A10BQwen125B1407.0575676~18Aug 2, 2026leaderboard64Gemini 3.1 Flash Lite PreviewGoogle1406.77391876Aug 2, 2026leaderboard65GLM 4.5Z.AI358B1404.6503638~6.2Aug 2, 2026leaderboard66Mistral Large 3 2512Mistral AI1403.1396935Aug 2, 2026leaderboard67GPT-4.1OpenAI1402.7334175Aug 2, 2026leaderboard68o3OpenAI1401.9307175Aug 2, 2026leaderboard69Gemini 2.5 FlashGoogle1400.96321121Aug 2, 2026leaderboard70Qwen3.5-27BQwen27.8B1400.4724898~80Aug 2, 2026leaderboard71Hy3 previewTencent299B1399.06796662~7.4Aug 2, 2026leaderboard72R1 0528DeepSeek685B1397.386650~3.2Aug 2, 2026leaderboard73DeepSeek V3.1 TerminusDeepSeek685B1393.43131467~3.2Aug 2, 2026leaderboard74Qwen3.5-35B-A3BQwen36B1388.63331389~61Aug 2, 2026leaderboard75GPT-5.4 NanoOpenAI1387.8079via High2220Aug 2, 2026leaderboard76Step 3.5 FlashStepFun199B1385.41954618~11Aug 2, 2026leaderboard77Qwen3 235B A22B Thinking 2507Qwen235B1385.4035927~9.3Aug 2, 2026leaderboard78Qwen3 Coder 480B A35BQwen480B1385.22041385~4.6Aug 2, 2026leaderboard79Qwen3 VL 235B A22B ThinkingQwen236B1383.1432via Thinking532~9.3Aug 2, 2026leaderboard80Qwen3.5-FlashQwen1382.36035317Aug 2, 2026leaderboard81MiniMax M2.5MiniMax229B1380.31381534~9.6Aug 2, 2026leaderboard82DeepSeek V3 0324DeepSeek685B1378.78251532~3.2Aug 2, 2026leaderboard83Qwen3 Next 80B A3B InstructQwen81.3B1378.56341767~27Aug 2, 2026leaderboard84GPT-5 MiniOpenAI1374.3798via High1374Aug 2, 2026leaderboard85GPT-4.1 MiniOpenAI1372.8154858Aug 2, 2026leaderboard86Claude Sonnet 5Anthropic1372.6912137Aug 2, 2026leaderboard87o4 MiniOpenAI1368.6203311Aug 2, 2026leaderboard88GLM 4.6VZ.AI108B1367.2371519~20Aug 2, 2026leaderboard89o3 Mini HighOpenAI1366.5149via High311Aug 2, 2026leaderboard90Qwen3 30B A3B Instruct 2507Qwen30.5B1366.39957078~71Aug 2, 2026leaderboard91GLM 4.5 AirZ.AI111B1362.22341603~20Aug 2, 2026leaderboard92Qwen3 Next 80B A3B ThinkingQwen81.3B1358.517via Thinking1742~27Aug 2, 2026leaderboard93Trinity Large ThinkingArcee AI399B1358.0734via Thinking1698~5.4Aug 2, 2026leaderboard94GLM 4.7 FlashZ.AI31.2B1350.3457337671Aug 2, 2026leaderboard95MiniMax M1MiniMax1346.1927612Aug 2, 2026leaderboard96Nemotron 3 SuperNVIDIA124B1345.63723364~17Aug 2, 2026leaderboard97Gemma 3 27BGoogle27.4B1343.87018399~78Aug 2, 2026leaderboard98DeepSeek V3DeepSeek685B1343.747515103Aug 2, 2026leaderboard99o3 MiniOpenAI1343.5408305Aug 2, 2026leaderboard100GLM 4.5VZ.AI108B1341.6523745~20Aug 2, 2026leaderboard

A tilde in Score / GB VRAM means the memory figure behind it is calculated from the parameter count rather than read from a published file. Those ratios are shown to two significant figures, and rows within a few percent of each other are not distinguishable.