Gemini 3.1 Pro Preview

Composite indexes

• Artificial Analysis Intelligence Index#47 of 151
  30independentv4.3.2, high thinkingArtificial Analysis: Gemini 3.1 Pro Preview ↗

  Leader: Claude Opus 5.5 at 58

Coding

• SciCode#13 of 100
  58.7%independentAA run, high thinkingArtificial Analysis: Gemini 3.1 Pro Preview ↗

  Leader: Claude Opus 5.5 at 66.9%

• Terminal-Bench 4.0#45 of 96
  4%independentAA run (Terminal-Bench 4.0), high thinkingArtificial Analysis: Gemini 3.1 Pro Preview ↗

  Leader: Claude Sonnet 5.5 at 70.6%

• Terminal-Bench 2.x#33 of 105
  73.8%independentAA run (Terminal-Bench 2.1), high thinkingArtificial Analysis: Gemini 3.1 Pro Preview ↗
  68.5%vendorTerminal-Bench 2.0, Terminus-2 harnessGoogle DeepMind: Gemini 3.1 Pro ↗

  Leader: Claude Fable 5.1 at 91.4%

• SWE-bench Pro#23 of 33
  54.2%vendorpublic set, single attempt, Thinking (High)Google DeepMind: Gemini 3.1 Pro ↗

  Leader: Claude Opus 5.5 at 89.9%

• SWE-bench Verified#3 of 18
  80.6%vendorsingle attempt, Thinking (High)Google DeepMind: Gemini 3.1 Pro ↗

  Leader: Claude Sonnet 5 at 85.2%

Agents and tool use

• GDPval-AA#69 of 109
  776independentGDPval-AA v2.1, high thinkingArtificial Analysis: Gemini 3.1 Pro Preview ↗

  Leader: Muse Wonder Full at 1857

• τ²-Bench#7 of 89
  99.3%vendortelecom (retail 90.8%)Google DeepMind: Gemini 3.1 Pro ↗
  95.6%independentAA run, telecom, high thinkingArtificial Analysis: Gemini 3.1 Pro Preview ↗

  Leader: GLM-5.2 at 99.1%

• BrowseComp#4 of 16
  85.9%vendorsearch + Python + browseGoogle DeepMind: Gemini 3.1 Pro ↗

  Leader: GPT-6 Astra at 91.5%

• MCP Atlas#9 of 10
  69.2%vendorGoogle DeepMind: Gemini 3.1 Pro ↗

  Leader: Kimi K3 at 84.2%

Reasoning and knowledge

• GPQA Diamond#9 of 144
  94.1%independentAA run, high thinkingArtificial Analysis: Gemini 3.1 Pro Preview ↗

  Leader: GPT-6 Astra at 96.1%

• Humanity’s Last Exam#17 of 153
  51.4%vendorsearch (blocklist) + codeGoogle DeepMind: Gemini 3.1 Pro ↗
  47%independentAA run, no tools, high thinkingArtificial Analysis: Gemini 3.1 Pro Preview ↗
  44.4%vendorno tools, Thinking (High)Google DeepMind: Gemini 3.1 Pro ↗

  Leader: Claude Sonnet 5.5 at 64.5%

• CritPt#26 of 144
  17.7%independentAA run, high thinkingArtificial Analysis: Gemini 3.1 Pro Preview ↗

  Leader: Muse Wonder Full at 33%

• ARC-AGI-2#10 of 25
  77.1%independentARC Prize verifiedARC Prize leaderboard data (semi-private) ↗

  Leader: GPT-6 Astra at 95%

Vision and long context

• AA Long Context Reasoning#21 of 144
  82%independentAA-LCR v1.1, high thinkingArtificial Analysis: Gemini 3.1 Pro Preview ↗

  Leader: Kimi K3 at 88.7%

• MMMU-Pro#11 of 72
  82.4%independentAA run, high thinkingArtificial Analysis: Gemini 3.1 Pro Preview ↗

  Leader: Claude Opus 5.5 at 87.7%

Human preference

• LMArena Text#10 of 57
  1487independentgemini-3.1-pro-preview; 106,951 votesLMArena text leaderboard ↗

  Leader: Gemini 4 Argon at 1525

• LMArena WebDev#47 of 58
  1447independentgemini-3.1-pro-preview; 22,889 votesLMArena Code Arena | WebDev leaderboard ↗

  Leader: Claude Opus 5.5 at 1818

Other published results