FRFrontier

THE FRONTIER GAP

Best available, release by release.

The best proprietary model against the best open-weight model on Artificial Analysis’ Intelligence Index, at every release that moved either line.

PROPRIETARY LEAD11.3 ptsClaude Opus 5.5 57.6 · MiMo-V2.6-Pro 46.3
PROPRIETARY TIME LEAD3.4 moClaude Fable 5 reached 46.3 before MiMo-V2.6-Pro

Artificial Analysis Intelligence Index v4.3.2: ten evaluations of reasoning, knowledge, coding, and agentic work. Refreshed September 26, 2026; AA estimates are marked.

Intelligence score ↑
Proprietary frontier Open-weight frontier Difference

The Intelligence history uses the September 26, 2026 snapshot of Artificial Analysis’ v4.3.2 index for every release, including older models; it does not mix scores from different index versions. At each date, the highest score available so far defines each frontier. The time gap finds the first model in the other group that matched the lower score. AA estimates many older scores rather than measuring them; those carry a * here. Coding and Agentic retain their separately labeled September 1 archive.

FRONTIER CHANGELOG

Releases that moved the line.

Key changes to the Intelligence Index frontier, newest first. * marks an Artificial Analysis estimate.

  1. PROPRIETARYClaude Opus 5.557.6
    Closed +11.3 pts
  2. OPEN WEIGHTSMiMo-V2.6-Pro46.3
    Closed +7.0 pts
  3. PROPRIETARYClaude Fable 5.153.4
    Closed +8.6 pts
  4. OPEN WEIGHTSGLM-5.344.8
    Closed +6.0 pts
  5. PROPRIETARYClaude Opus 550.8
    Closed +7.2 pts
  6. OPEN WEIGHTSKimi K343.6
    Closed +6.0 pts
  7. OPEN WEIGHTSGLM-5.233.7
    Closed +15.9 pts
  8. PROPRIETARYClaude Fable 549.6
    Closed +20.4 pts
  9. OPEN WEIGHTSMiniMax M329.2
    Closed +12.6 pts
  10. PROPRIETARYClaude Opus 4.841.8
    Closed +17.6 pts
  11. OPEN WEIGHTSDeepSeek V4 Flash24.2
    Closed +16.5 pts
  12. PROPRIETARYClaude Opus 4.740.7*
    Closed +17.8 pts
  13. PROPRIETARYGPT-5.439*
    Closed +16.1 pts
  14. OPEN WEIGHTSQwen3.5 27B22.9*
    Closed +7.6 pts
  15. OPEN WEIGHTSGLM-4.7 Reasoning22.2*
    Closed +8.2 pts
  16. PROPRIETARYGPT-5.230.4*
    Closed +8.4 pts
  17. PROPRIETARYGemini 3 Pro (Preview)28*
    Closed +5.9 pts
  18. PROPRIETARYGPT-5.124.7*
    Closed +2.7 pts
  19. OPEN WEIGHTSKimi K2 Thinking22*
    Closed +1.0 pts
  20. OPEN WEIGHTSGLM-4.6 Reasoning18.5*
    Closed +4.4 pts
  21. PROPRIETARYGPT-523*
    Closed +9.9 pts
  22. OPEN WEIGHTSDeepSeek R1 052813.1*
    Closed +7.1 pts
  23. PROPRIETARYo320.2*
    Closed +8.8 pts
  24. PROPRIETARYClaude 3.7 Sonnet15.3*
    Closed +3.9 pts
  25. OPEN WEIGHTSDeepSeek R1 (Jan ’25)11.4
    Closed +3.8 pts
  26. OPEN WEIGHTSDeepSeek V3 (Dec ’24)8.5
    Closed +6.7 pts
  27. PROPRIETARYo115.2*
    Closed +7.5 pts
  28. OPEN WEIGHTSQwen2.5 72B7.7*
    Closed +3.7 pts
  29. PROPRIETARYo1-preview11.4*
    Closed +4.1 pts
  30. OPEN WEIGHTSLlama 3.1 405B7.3*
    Closed +1.4 pts
  31. PROPRIETARYClaude 3 Opus8.7*OPEN WEIGHTSMixtral 8x22B5.7*
    Closed +3.0 pts
SourcesIntelligence Index methodology ↗Artificial Analysis model pages ↗

Benchmark data © Artificial Analysis and used with attribution. Each score is read from the model’s Artificial Analysis page; a model without a published score is left off the chart, never estimated.