Mercury 2.5

Composite indexes

• Artificial Analysis Intelligence Index#96 of 151
  12independentIntelligence Index v4.3; reasoningArtificial Analysis: Mercury 2.5 ↗

  Leader: Claude Opus 5.5 at 58

Coding

• SciCode#82 of 100
  38.5%independentreasoningArtificial Analysis: Mercury 2.5 ↗

  Leader: Claude Opus 5.5 at 66.9%

• Terminal-Bench 4.0#96 of 96
  0%independentreasoningArtificial Analysis: Mercury 2.5 ↗

  Leader: Claude Sonnet 5.5 at 70.6%

Agents and tool use

• GDPval-AA#85 of 109
  462independentreasoningArtificial Analysis: Mercury 2.5 ↗

  Leader: Muse Wonder Full at 1857

Reasoning and knowledge

• Humanity’s Last Exam#96 of 153
  11.8%independentno tools; reasoningArtificial Analysis: Mercury 2.5 ↗

  Leader: Claude Sonnet 5.5 at 64.5%

• CritPt#140 of 144
  0%independentreasoningArtificial Analysis: Mercury 2.5 ↗

  Leader: Muse Wonder Full at 33%

Vision and long context

• AA Long Context Reasoning#77 of 144
  71.7%independentreasoningArtificial Analysis: Mercury 2.5 ↗

  Leader: Kimi K3 at 88.7%

Other published results