Leaderboards
Benchmarks
Every score carries its origin — leaderboard is independent,vendor is self-reported. History is kept, never overwritten.
Arena Creative Writing143Arena Instruction Following143Arena Text (overall)143IFEval16LiveBench Instruction Following35LiveBench Language35
Arena Text (overall)
Arena Score (Bradley-Terry, style-controlled) from human pairwise votes — not a percentagehigher is betterProvenanceindependent leaderboard100 modelsscores Aug 2, 2026
1Claude Fable 5Anthropic1508.576930—Aug 2, 2026leaderboard2Claude Opus 4.6Anthropic1497.188960—Aug 2, 2026leaderboard3Claude Opus 4.7Anthropic1492.408960—Aug 2, 2026leaderboard4Claude Opus 5Anthropic1491.8152via High60—Aug 2, 2026leaderboard5Muse Spark 1.1Meta1489.6718351—Aug 2, 2026leaderboard6Gemini 3.1 Pro Preview Custom ToolsGoogle1485.4401248—Aug 2, 2026leaderboard7Kimi K3Moonshot AI2.8T1485.3457via Max106~0.85Aug 2, 2026leaderboard8GPT-5.6 SolOpenAI1482.7715via xHigh99—Aug 2, 2026leaderboard9Gemini 3.6 FlashGoogle1482.6824395—Aug 2, 2026leaderboard10Gemini 3.5 FlashGoogle1476.3958via High328—Aug 2, 2026leaderboard11GPT-5.5OpenAI1476.3335148—Aug 2, 2026leaderboard12GPT-5.2 ChatOpenAI1476.1225211—Aug 2, 2026leaderboard13Qwen3.7 MaxQwen1474.6573393—Aug 2, 2026leaderboard14Claude Opus 4.8Anthropic1474.574859—Aug 2, 2026leaderboard15Grok 4.20xAI1474.3776590—Aug 2, 2026leaderboard16Claude Sonnet 4.6Anthropic1471.737198—Aug 2, 2026leaderboard17Grok 4.20 Multi-AgentxAI1470.8908588—Aug 2, 2026leaderboard18GLM 5.2Z.AI753B1469.4933via Max980~3.1Aug 2, 2026leaderboard19Claude Opus 4.5Anthropic1468.866859—Aug 2, 2026leaderboard20Grok 4.5xAI1468.8266245—Aug 2, 2026leaderboard21GLM 5.1Z.AI754B1468.7878491~3.1Aug 2, 2026leaderboard22GPT-5.6 TerraOpenAI1467.8148via xHigh489—Aug 2, 2026leaderboard23MiMo-V2.5-ProXiaomi1T1465.70792106~2.3Aug 2, 2026leaderboard24GPT-5.4OpenAI1465.3677195—Aug 2, 2026leaderboard25Kimi K2.6Moonshot AI1.1T1460.668589~2.2Aug 2, 2026leaderboard26Qwen3.6 Max PreviewQwen1459.9104237—Aug 2, 2026leaderboard27Qwen3.7 PlusQwen1459.431140—Aug 2, 2026leaderboard28DeepSeek V4 ProDeepSeek1.6T1457.94521676~1.4Aug 2, 2026leaderboard29GLM 5Z.AI754B1456.7132759~3.1Aug 2, 2026leaderboard30Gemini 3.5 Flash LiteGoogle1456.50081165—Aug 2, 2026leaderboard31Hy3Tencent299B1456.27282758~7.7Aug 2, 2026leaderboard32Claude Sonnet 4.5Anthropic1454.965197—Aug 2, 2026leaderboard33Gemma 4 31BGoogle32.7B1450.92334267~70Aug 2, 2026leaderboard34GPT-5.6 LunaOpenAI1450.403via xHigh1099—Aug 2, 2026leaderboard35Kimi K2 ThinkingMoonshot AI1.1T1450.212via Thinking580~2.2Aug 2, 2026leaderboard36GPT-5.3 ChatOpenAI1448.6971103—Aug 2, 2026leaderboard37GPT-5.4 MiniOpenAI1448.6329via High644—Aug 2, 2026leaderboard38Claude Opus 4.1Anthropic1447.114519—Aug 2, 2026leaderboard39Gemini 2.5 ProGoogle1445.533289—Aug 2, 2026leaderboard40MiniMax M3MiniMax427B1445.27321505~5.4Aug 2, 2026leaderboard41Qwen3.6 PlusQwen1443.5591740—Aug 2, 2026leaderboard42Grok 4.3xAI1442.3961577—Aug 2, 2026leaderboard43GLM 4.7Z.AI358B1442.0523824~6.4Aug 2, 2026leaderboard44Qwen3.5 397B A17BQwen403B1441.6808616~5.7Aug 2, 2026leaderboard45InklingThinking Machines952B1440.8017356~2.4Aug 2, 2026leaderboard46GPT-5.1OpenAI1438.6689288—Aug 2, 2026leaderboard47Gemma 4 26B A4BGoogle26.5B1438.49654795~86Aug 2, 2026leaderboard48DeepSeek V4 FlashDeepSeek291B1435.63628139~7.8Aug 2, 2026leaderboard49Qwen3 MaxQwen1434.6103368—Aug 2, 2026leaderboard50MiMo-V2.5Xiaomi311B1433.73856401~7.3Aug 2, 2026leaderboard51GLM 5V TurboZ.AI1432.9843358—Aug 2, 2026leaderboard52Gemini 3.1 Flash Lite PreviewGoogle1432.38961910—Aug 2, 2026leaderboard53Kimi K2.5Moonshot AI1.1T1431.5938716~2.1Aug 2, 2026leaderboard54o3OpenAI1430.9088179—Aug 2, 2026leaderboard55Mistral Medium 3.5Mistral AI128B1427.2916190~18Aug 2, 2026leaderboard56DeepSeek V3.2DeepSeek685B1425.04194585~3.3Aug 2, 2026leaderboard57GLM 4.6Z.AI357B1425.0247814~6.3Aug 2, 2026leaderboard58Qwen3 235B A22B Instruct 2507Qwen235B1422.94932587~9.6Aug 2, 2026leaderboard59DeepSeek V3.2 ExpDeepSeek685B1422.62753470~3.3Aug 2, 2026leaderboard60R1 0528DeepSeek685B1422.1116661~3.3Aug 2, 2026leaderboard61MiniMax M2.7MiniMax229B1416.88151476~9.8Aug 2, 2026leaderboard62Qwen3.5-122B-A10BQwen125B1416.8699681~18Aug 2, 2026leaderboard63Mistral Large 3 2512Mistral AI1415.4549944—Aug 2, 2026leaderboard64Qwen3 VL 235B A22B InstructQwen236B1415.12541608~9.5Aug 2, 2026leaderboard65DeepSeek V3.1 TerminusDeepSeek685B1415.07381490~3.3Aug 2, 2026leaderboard66GPT-4.1OpenAI1413.7743177—Aug 2, 2026leaderboard67Claude Opus 4Anthropic1412.725419—Aug 2, 2026leaderboard68Claude Haiku 4.5Anthropic1412.2344282—Aug 2, 2026leaderboard69Hy3 previewTencent299B1412.19046725~7.5Aug 2, 2026leaderboard70GLM 4.5Z.AI358B1410.7141641~6.2Aug 2, 2026leaderboard71Gemini 2.5 FlashGoogle1410.23971128—Aug 2, 2026leaderboard72Qwen3.5-27BQwen27.8B1408.0397903~80Aug 2, 2026leaderboard73GPT-5.4 NanoOpenAI1402.385via High2244—Aug 2, 2026leaderboard74o1OpenAI1402.169723—Aug 2, 2026leaderboard75Qwen3 Next 80B A3B InstructQwen81.3B1401.20751796~27Aug 2, 2026leaderboard76Qwen3 235B A22B Thinking 2507Qwen235B1398.9191936~9.4Aug 2, 2026leaderboard77Qwen3.5-FlashQwen1396.81865372—Aug 2, 2026leaderboard78DeepSeek V3 0324DeepSeek685B1395.53151551~3.2Aug 2, 2026leaderboard79Qwen3.5-35B-A3BQwen36B1395.49471395~61Aug 2, 2026leaderboard80Qwen3 VL 235B A22B ThinkingQwen236B1395.3412via Thinking537~9.4Aug 2, 2026leaderboard81Step 3.5 FlashStepFun199B1394.60754649~11Aug 2, 2026leaderboard82GPT-5 MiniOpenAI1390.043via High1390—Aug 2, 2026leaderboard83o4 MiniOpenAI1390.0244316—Aug 2, 2026leaderboard84MiniMax M2.5MiniMax229B1389.92871544~9.6Aug 2, 2026leaderboard85Qwen3 Coder 480B A35BQwen480B1387.71031388~4.6Aug 2, 2026leaderboard86Qwen3 30B A3B Instruct 2507Qwen30.5B1383.09877164~72Aug 2, 2026leaderboard87GPT-4.1 MiniOpenAI1382.9728864—Aug 2, 2026leaderboard88GLM 4.6VZ.AI108B1377.23331530~20Aug 2, 2026leaderboard89Claude Sonnet 5Anthropic1373.3222137—Aug 2, 2026leaderboard90GLM 4.5 AirZ.AI111B1372.74681615~20Aug 2, 2026leaderboard91Qwen3 Next 80B A3B ThinkingQwen81.3B1369.3246via Thinking1756~27Aug 2, 2026leaderboard92Trinity Large ThinkingArcee AI399B1368.892via Thinking1711~5.4Aug 2, 2026leaderboard93GLM 4.7 FlashZ.AI31.2B1367.826342072Aug 2, 2026leaderboard94Gemma 3 27BGoogle27.4B1365.90478537~79Aug 2, 2026leaderboard95MiniMax M1MiniMax1363.7725620—Aug 2, 2026leaderboard96o3 Mini HighOpenAI1363.554via High310—Aug 2, 2026leaderboard97Nemotron 3 SuperNVIDIA124B1360.84733402~17Aug 2, 2026leaderboard98DeepSeek V3DeepSeek685B1358.543715263Aug 2, 2026leaderboard99GLM 4.5VZ.AI108B1353.3372752~20Aug 2, 2026leaderboard100gpt-oss-120bOpenAI120B1352.29337955~18Aug 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.