Kimi K3

Composite indexes

• Artificial Analysis Intelligence Index#14 of 142
  44independentv4.3.2, reasoning, max effortArtificial Analysis: Kimi K3 (max) ↗
  34independentv4.3.2, reasoning, low effort (AA-estimated)Artificial Analysis: Kimi K3 (low) ↗

  Leader: Claude Opus 5.5 at 58

Coding

• Terminal-Bench 4.0#27 of 88
  12.6%independentreasoning, max effort, AA-runArtificial Analysis: Kimi K3 (max) ↗

  Leader: Claude Mythos 5.1 at 60.9%

• Terminal-Bench 2.x#15 of 103
  88.3%vendorTerminal-Bench 2.1, maxKimi K3 model card (Hugging Face) ↗
  85%independentTerminal-Bench 2.1, reasoning, max effort, AA-runArtificial Analysis: Kimi K3 (max) ↗

  Leader: Claude Fable 5.1 at 91.4%

• SciCode#6 of 94
  59.5%independentreasoning, max effort, AA-runArtificial Analysis: Kimi K3 (max) ↗
  58.7%vendormaxKimi K3 model card (Hugging Face) ↗

  Leader: Claude Opus 5.5 at 66.9%

Agents and tool use

• GDPval-AA#19 of 103
  1686vendorGDPval-AA v2, max (as reported by Moonshot)Kimi K3 model card (Hugging Face) ↗
  1524independentGDPval-AA v2.1, reasoning, max effortArtificial Analysis: Kimi K3 (max) ↗

  Leader: Claude Opus 5.5 at 1846

• BrowseComp#2 of 14
  91.2%vendormaxKimi K3 model card (Hugging Face) ↗

  Leader: GPT-6 Astra at 91.5%

• MCP Atlas#1 of 9
  84.2%vendormaxKimi K3 model card (Hugging Face) ↗

  Leader: Kimi K3 at 84.2%

• OSWorld-Verified#1 of 16
  84.8%vendorOSWorld-Verified, maxKimi K3 model card (Hugging Face) ↗

  Leader: Kimi K3 at 84.8%

Reasoning and knowledge

• GPQA Diamond#13 of 142
  93.5%independentreasoning, max effort, AA-runArtificial Analysis: Kimi K3 (max) ↗
  93.5%vendormaxKimi K3 model card (Hugging Face) ↗

  Leader: GPT-6 Astra at 96.1%

• Humanity’s Last Exam#16 of 147
  56%vendormax, with tools, full setKimi K3 model card (Hugging Face) ↗
  46.9%independentreasoning, max effort, no tools, AA-runArtificial Analysis: Kimi K3 (max) ↗
  43.5%vendormax, no tools, full setKimi K3 model card (Hugging Face) ↗

  Leader: Claude Opus 5.5 at 61.4%

• CritPt#11 of 139
  23.4%independentreasoning, max effort, AA-runArtificial Analysis: Kimi K3 (max) ↗
  23.4%vendormaxKimi K3 model card (Hugging Face) ↗

  Leader: GPT-5.6 Sol at 32.3%

Vision and long context

• AA Long Context Reasoning#1 of 139
  88.7%independentreasoning, max effort, AA-LCRArtificial Analysis: Kimi K3 (max) ↗
  74.7%vendormax (vendor-run)Kimi K3 model card (Hugging Face) ↗

  Leader: Kimi K3 at 88.7%

• MMMU-Pro#15 of 72
  80.5%independentreasoning, max effort, AA-runArtificial Analysis: Kimi K3 (max) ↗

  Leader: Claude Opus 5.5 at 87.7%

• Video-MME#1 of 4
  90%vendorwith subtitles, maxKimi K3 model card (Hugging Face) ↗

  Leader: Kimi K3 at 90%

Human preference

• LMArena Text#10 of 54
  1485independentarena entry "kimi-k3-max", 20,987 votes, checked 2026-09-25Arena (LMArena) text leaderboard ↗

  Leader: Claude Fable 5 at 1506

• LMArena WebDev#7 of 54
  1660independentarena entry "kimi-k3-max", 13,252 votes, checked 2026-09-25Arena (LMArena) WebDev/Code leaderboard ↗

  Leader: Claude Opus 5.5 at 1818

Other published results