super-intelligence.compare
SOURCEDArena human preference · text / overallMethod & sources ↗Checked 12 Oct 2026

Every model, sized by what you care about.

gemini-4-argon-high · 1533.9 ratinggemini 4argon high1534 rating1claude-opus-5.5-high · 1514.8 ratingclaude opus5.5 high1515 rating2claude-fable-5.1-max · 1510.0 ratingclaude fable5.1 max1510 rating3claude-opus-5-max · 1507.0 ratingclaude opus5 max1507 ratingclaude-opus-4-6-high · 1502.9 ratingclaude opus4 6 high1503 ratingclaude-opus-5-high · 1502.6 ratingclaude opus5 high1503 ratinggemini-3.8-flash-high · 1499.0 ratinggemini 3.8flash high1499 ratingclaude-opus-4-6 · 1498.1 ratingclaude opus4 61498 ratingmimo-v2.6-pro · 1491.7 ratingmimo v2.6pro1492 ratingclaude-fable-5-high · 1491.2 ratingclaude fable5 high1491 ratingclaude-opus-4-7-high · 1489.6 ratingclaude opus4 7 high1490 ratingmuse-spark-1.3-max · 1489.6 ratingmuse spark1.3 max1490 ratinggemini-3.7-flash-high · 1486.0 ratinggemini 3.7flash high1486 ratingqwen3.8-max · 1483.3 ratingqwen3.8 max1483 ratingclaude-opus-4-7 · 1483.2 ratingclaude opus4 71483 ratingmuse-spark-1.2 (xHigh) · 1482.3 ratingmuse spark1.2 (xHigh)1482 ratinggemini-3.5-flash-high · 1481.9 ratinggemini 3.5flash high1482 ratinggemini-3.1-pro-preview · 1480.5 ratinggemini 3.1pro preview1481 ratingglm-5.3-max · 1471.2 ratingglm 5.3 max1471 ratingglm-5.3-flash · 1471.1 ratingglm 5.3flash1471 ratingdeepseek-v4-pro · 1451.1 ratingdeepseekv4 pro1451 ratingSOURCED · Human preference; rating scale, not a 0–100 intelligence score.
GoogleAnthropicXiaomiMetaAlibabazaiDeepSeek
Board published · 8 Oct 2026