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 Creative Writing
Arena Score on creative-writing prompts — human preference, not a prose-craft rubrichigher is betterProvenanceindependent leaderboard100 modelsscores Aug 2, 2026
1Claude Fable 5Anthropic1506.639430—Aug 2, 2026leaderboard2Claude Opus 4.7Anthropic1480.036259—Aug 2, 2026leaderboard3Gemini 3.1 Pro Preview Custom ToolsGoogle1479.0136247—Aug 2, 2026leaderboard4Claude Opus 4.6Anthropic1478.013359—Aug 2, 2026leaderboard5Claude Opus 5Anthropic1477.3813via Max59—Aug 2, 2026leaderboard6GPT-5.6 SolOpenAI1475.5558via xHigh98—Aug 2, 2026leaderboard7Kimi K3Moonshot AI2.8T1468.7657via Max105~0.84Aug 2, 2026leaderboard8Gemini 3.5 FlashGoogle1463.6849via High325—Aug 2, 2026leaderboard9Muse Spark 1.1Meta1462.9995344—Aug 2, 2026leaderboard10Gemini 3.6 FlashGoogle1462.7566390—Aug 2, 2026leaderboard11Claude Opus 4.8Anthropic1462.143358—Aug 2, 2026leaderboard12Grok 4.20xAI1461.1425584—Aug 2, 2026leaderboard13Claude Opus 4.5Anthropic1459.97358—Aug 2, 2026leaderboard14GPT-5.5OpenAI1456.6027146—Aug 2, 2026leaderboard15Claude Sonnet 4.5Anthropic1453.214697—Aug 2, 2026leaderboard16Claude Sonnet 4.6Anthropic1451.371497—Aug 2, 2026leaderboard17Grok 4.20 Multi-AgentxAI1448.4355579—Aug 2, 2026leaderboard18GLM 5.1Z.AI754B1448.0723484~3Aug 2, 2026leaderboard19Grok 4.5xAI1447.6443241—Aug 2, 2026leaderboard20Qwen3.7 MaxQwen1446.8386—Aug 2, 2026leaderboard21GLM 5Z.AI754B1446.2097753~3Aug 2, 2026leaderboard22DeepSeek V4 ProDeepSeek1.6T1446.13021662~1.4Aug 2, 2026leaderboard23GLM 5.2Z.AI753B1444.5809via Max963~3Aug 2, 2026leaderboard24Gemini 2.5 ProGoogle1444.1422289—Aug 2, 2026leaderboard25Claude Opus 4.1Anthropic1441.921619—Aug 2, 2026leaderboard26Qwen3.7 PlusQwen1436.71371122—Aug 2, 2026leaderboard27Qwen3.6 Max PreviewQwen1436.5216233—Aug 2, 2026leaderboard28Gemini 3.5 Flash LiteGoogle1434.49311148—Aug 2, 2026leaderboard29GPT-5.4OpenAI1434.2422191—Aug 2, 2026leaderboard30GPT-5.2 ChatOpenAI1433.2418205—Aug 2, 2026leaderboard31MiMo-V2.5-ProXiaomi1T1431.09572056~2.2Aug 2, 2026leaderboard32Kimi K2.6Moonshot AI1.1T1429.4789576~2.1Aug 2, 2026leaderboard33Grok 4.3xAI1427.653571—Aug 2, 2026leaderboard34Hy3Tencent299B1424.0872697~7.6Aug 2, 2026leaderboard35Kimi K2 ThinkingMoonshot AI1.1T1423.2574via Thinking569~2.1Aug 2, 2026leaderboard36Gemma 4 31BGoogle32.7B1420.56464178~69Aug 2, 2026leaderboard37GPT-5.6 TerraOpenAI1418.7485via xHigh473—Aug 2, 2026leaderboard38GPT-5.6 LunaOpenAI1417.8577via xHigh1074—Aug 2, 2026leaderboard39Claude Opus 4Anthropic1416.767419—Aug 2, 2026leaderboard40Gemini 3.1 Flash Lite PreviewGoogle1413.4731885—Aug 2, 2026leaderboard41GPT-5.1OpenAI1410.1011282—Aug 2, 2026leaderboard42DeepSeek V3.2 ExpDeepSeek685B1409.98553439~3.3Aug 2, 2026leaderboard43DeepSeek V4 FlashDeepSeek291B1407.38077978~7.7Aug 2, 2026leaderboard44Qwen3.6 PlusQwen1406.8534721—Aug 2, 2026leaderboard45DeepSeek V3.1 TerminusDeepSeek685B1406.80121481~3.3Aug 2, 2026leaderboard46Qwen3.5 397B A17BQwen403B1406.7816601~5.5Aug 2, 2026leaderboard47GPT-5.3 ChatOpenAI1406.6361100—Aug 2, 2026leaderboard48MiniMax M3MiniMax427B1405.96671465~5.2Aug 2, 2026leaderboard49GLM 4.7Z.AI358B1405.0684803~6.2Aug 2, 2026leaderboard50Gemma 4 26B A4BGoogle26.5B1403.97924680~84Aug 2, 2026leaderboard51GPT-5.4 MiniOpenAI1402.869via High623—Aug 2, 2026leaderboard52GPT-4.1OpenAI1402.3607175—Aug 2, 2026leaderboard53GLM 4.6Z.AI357B1402.1588801~6.2Aug 2, 2026leaderboard54DeepSeek V3.2DeepSeek685B1400.37384506~3.2Aug 2, 2026leaderboard55GLM 5V TurboZ.AI1398.9807350—Aug 2, 2026leaderboard56Gemini 2.5 FlashGoogle1396.56321117—Aug 2, 2026leaderboard57Qwen3 MaxQwen1394.8258358—Aug 2, 2026leaderboard58Mistral Medium 3.5Mistral AI128B1394.4658186~17Aug 2, 2026leaderboard59MiMo-V2.5Xiaomi311B1394.10856224~7.1Aug 2, 2026leaderboard60R1 0528DeepSeek685B1393.5859648~3.2Aug 2, 2026leaderboard61DeepSeek V3 0324DeepSeek685B1390.07591545~3.2Aug 2, 2026leaderboard62Kimi K2.5Moonshot AI1.1T1389.3667695~2.1Aug 2, 2026leaderboard63Claude Haiku 4.5Anthropic1389.0351278—Aug 2, 2026leaderboard64InklingThinking Machines952B1384.4044342~2.3Aug 2, 2026leaderboard65o3OpenAI1383.1141173—Aug 2, 2026leaderboard66o1OpenAI1381.422323—Aug 2, 2026leaderboard67Qwen3 235B A22B Instruct 2507Qwen235B1379.98942509~9.3Aug 2, 2026leaderboard68Mistral Large 3 2512Mistral AI1376.2502918—Aug 2, 2026leaderboard69Qwen3 235B A22B Thinking 2507Qwen235B1373.3138919~9.3Aug 2, 2026leaderboard70GLM 4.5Z.AI358B1372.9662624~6.1Aug 2, 2026leaderboard71Claude Sonnet 5Anthropic1369.7557137—Aug 2, 2026leaderboard72Qwen3.5-122B-A10BQwen125B1367.0728657~17Aug 2, 2026leaderboard73Qwen3 Coder 480B A35BQwen480B1365.75581366~4.5Aug 2, 2026leaderboard74MiniMax M2.7MiniMax229B1365.52291422~9.5Aug 2, 2026leaderboard75Qwen3 VL 235B A22B InstructQwen236B1361.00871547~9.2Aug 2, 2026leaderboard76MiniMax M2.5MiniMax229B1357.27771508~9.4Aug 2, 2026leaderboard77Qwen3.5-27BQwen27.8B1357.1482870~77Aug 2, 2026leaderboard78Hy3 previewTencent299B1355.9156457~7.2Aug 2, 2026leaderboard79GPT-4.1 MiniOpenAI1349.6705844—Aug 2, 2026leaderboard80DeepSeek V3DeepSeek685B1348.899815163Aug 2, 2026leaderboard81Gemma 3 27BGoogle27.4B1347.6498423~78Aug 2, 2026leaderboard82Step 3.5 FlashStepFun199B1344.58764482~11Aug 2, 2026leaderboard83GLM 4.6VZ.AI108B1344.27351494~20Aug 2, 2026leaderboard84Qwen3.5-35B-A3BQwen36B1343.42911343~59Aug 2, 2026leaderboard85o4 MiniOpenAI1337.9415304—Aug 2, 2026leaderboard86GPT-5.4 NanoOpenAI1337.9186via High2141—Aug 2, 2026leaderboard87Qwen3 VL 235B A22B ThinkingQwen236B1337.7942via Thinking515~9Aug 2, 2026leaderboard88GPT-4o (2024-05-13)OpenAI1337.441589—Aug 2, 2026leaderboard89Qwen3.5-FlashQwen1337.36915144—Aug 2, 2026leaderboard90Trinity Large ThinkingArcee AI399B1334.0927via Thinking1668~5.3Aug 2, 2026leaderboard91Gemma 3 12BGoogle12.2B1333.55713336~170Aug 2, 2026leaderboard92GLM 4.5 AirZ.AI111B1327.41731562~19Aug 2, 2026leaderboard93GPT-4o (2024-08-06)OpenAI1326.7667133—Aug 2, 2026leaderboard94GPT-5 MiniOpenAI1324.93via High1325—Aug 2, 2026leaderboard95Qwen3 Next 80B A3B ThinkingQwen81.3B1323.697via Thinking1697~26Aug 2, 2026leaderboard96GPT-4 TurboOpenAI1322.815444—Aug 2, 2026leaderboard97Qwen3 30B A3B Instruct 2507Qwen30.5B1320.69026841~69Aug 2, 2026leaderboard98MiniMax M1MiniMax1317.517599—Aug 2, 2026leaderboard99Qwen3 Next 80B A3B InstructQwen81.3B1315.79821687~26Aug 2, 2026leaderboard100GLM 4.7 FlashZ.AI31.2B1312.0243328069Aug 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.