GPT-6 Luna

Composite indexes

• Artificial Analysis Intelligence Index#27 of 142
  37independentAA Intelligence Index v4.3.2, max effortArtificial Analysis: GPT-6 Luna (max) ↗

  Leader: Claude Opus 5.5 at 58

Coding

• Terminal-Bench 4.0#26 of 88
  12.6%independentTerminal-Bench 4.0 (AA harness), max effortArtificial Analysis: GPT-6 Luna (max) ↗

  Leader: Claude Mythos 5.1 at 60.9%

• SciCode#23 of 94
  54.6%independentmax effortArtificial Analysis: GPT-6 Luna (max) ↗

  Leader: Claude Opus 5.5 at 66.9%

Agents and tool use

• GDPval-AA#32 of 103
  1367independentGDPval-AA, max effortArtificial Analysis: GPT-6 Luna (max) ↗

  Leader: Claude Opus 5.5 at 1846

Reasoning and knowledge

• Humanity’s Last Exam#40 of 147
  38.5%independentno tools (AA), max effortArtificial Analysis: GPT-6 Luna (max) ↗

  Leader: Claude Opus 5.5 at 61.4%

• CritPt#17 of 139
  19.4%independentmax effortArtificial Analysis: GPT-6 Luna (max) ↗

  Leader: GPT-5.6 Sol at 32.3%

• ARC-AGI-2#16 of 25
  59.3%independentsemi-private, Max, $0.062/taskARC Prize leaderboard ↗

  Leader: GPT-6 Astra at 95%

• ARC-AGI-3#9 of 9
  0.1%independentsemi-private, MaxARC Prize leaderboard ↗

  Leader: GPT-6 Astra at 62.7%

Vision and long context

• AA Long Context Reasoning#10 of 139
  83.3%independentmax effortArtificial Analysis: GPT-6 Luna (max) ↗

  Leader: Kimi K3 at 88.7%

• MMMU-Pro#33 of 72
  75.5%independentAA, max effortArtificial Analysis: GPT-6 Luna (max) ↗

  Leader: Claude Opus 5.5 at 87.7%

Human preference

• LMArena WebDev#21 of 54
  1593independentgpt-6-luna-max, LMArena Code arena (formerly WebDev), vote cutoff 2026-09-23LMArena Code (WebDev) leaderboard ↗

  Leader: Claude Opus 5.5 at 1818

Other published results