Grok 4.5

Composite indexes

• Artificial Analysis Intelligence Index#32 of 151
  39independentAA Intelligence Index v4.3.2, highArtificial Analysis: Grok 4.5 (high) ↗

  Leader: Claude Opus 5.5 at 58

Coding

• Terminal-Bench 4.0#39 of 96
  10.6%independentTerminal-Bench 4.0 (AA harness), highArtificial Analysis: Grok 4.5 (high) ↗

  Leader: Claude Sonnet 5.5 at 70.6%

• SciCode#24 of 100
  55%independenthighArtificial Analysis: Grok 4.5 (high) ↗

  Leader: Claude Opus 5.5 at 66.9%

Agents and tool use

• GDPval-AA#36 of 109
  1370independentGDPval-AA, highArtificial Analysis: Grok 4.5 (high) ↗

  Leader: Muse Wonder Full at 1857

Reasoning and knowledge

• Humanity’s Last Exam#28 of 153
  42.7%independentno tools (AA), highArtificial Analysis: Grok 4.5 (high) ↗

  Leader: Claude Sonnet 5.5 at 64.5%

• GPQA Diamond#15 of 144
  93.1%independentAA, highArtificial Analysis: Grok 4.5 (high) ↗

  Leader: GPT-6 Astra at 96.1%

• CritPt#33 of 144
  15.4%independenthighArtificial Analysis: Grok 4.5 (high) ↗

  Leader: Muse Wonder Full at 33%

• ARC-AGI-2#17 of 25
  52.6%independentsemi-private, HighARC Prize leaderboard ↗

  Leader: GPT-6 Astra at 95%

Vision and long context

• AA Long Context Reasoning#41 of 144
  79.3%independenthighArtificial Analysis: Grok 4.5 (high) ↗

  Leader: Kimi K3 at 88.7%

• MMMU-Pro#16 of 72
  80.4%independentAA, highArtificial Analysis: Grok 4.5 (high) ↗

  Leader: Claude Opus 5.5 at 87.7%

Human preference

• LMArena Text#22 of 57
  1468independentgrok-4.5, style control default, vote cutoff 2026-09-13LMArena Text leaderboard ↗

  Leader: Gemini 4 Argon at 1525

• LMArena WebDev#30 of 58
  1552independentgrok-4.5, LMArena Code arena (formerly WebDev), vote cutoff 2026-09-23LMArena Code (WebDev) leaderboard ↗

  Leader: Claude Opus 5.5 at 1818

Other published results