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

Archive reconstruction: publisher-labelled history may contain later restatements. Not a backtest. Loading full history…

Who’s keeping their head above water?

Above water = a higher preference rating than the strongest open-weight model on this board.

8 above / 42 at or below
OPEN-WEIGHT BASELINE · mimo v2.6 pro · 1491.7 rating-400-200-1000+100+200+400gemini 4argon high+42 ratinggemini-4-argon-high · 42.2 rating · Closed weightsclaude opus5.5 high+23 ratingclaude-opus-5.5-high · 23.2 rating · Closed weightsclaude fable5.1 max+18 ratingclaude-fable-5.1-max · 18.4 rating · Closed weightsclaude opus5 max+15 ratingclaude-opus-5-max · 15.3 rating · Closed weightsclaude opus4 6 high+11 ratingclaude-opus-4-6-high · 11.2 rating · Closed weightsclaude opus5 high+11 ratingclaude-opus-5-high · 10.9 rating · Closed weightsgemini 3.8flash high+7 ratinggemini-3.8-flash-high · 7.3 rating · Closed weightsclaude opus4 6+6 ratingclaude-opus-4-6 · 6.4 rating · Closed weightsmimo v2.6pro0 ratingmimo-v2.6-pro · 0.0 rating · Open weightsclaude fable5 high-0 ratingclaude-fable-5-high · -0.4 rating · Closed weightsclaude opus4 7 high-2 ratingclaude-opus-4-7-high · -2.0 rating · Closed weightsmuse spark1.3 max-2 ratingmuse-spark-1.3-max · -2.1 rating · Closed weightsgemini 3.7flash high-6 ratinggemini-3.7-flash-high · -5.6 rating · Closed weightsqwen3.8 max-8 ratingqwen3.8-max · -8.3 rating · Closed weightsclaude opus4 7-8 ratingclaude-opus-4-7 · -8.4 rating · Closed weightsmuse-spark-1.2 (xHigh) · -9.4 rating · Closed weightsgemini-3.5-flash-high · -9.7 rating · Closed weightsgemini-3.1-pro-preview · -11.1 rating · Closed weightsgemini-3-pro · -12.3 rating · Closed weightsgemini-3.6-flash-high · -12.9 rating · Closed weightsmuse-spark-1.1 · -13.1 rating · Closed weightsgemini-3.5-flash-medium · -14.2 rating · Closed weightsqwen3.7-max-preview · -16.0 rating · Closed weightskimi-k3-max · -16.2 rating · Weights unknownmuse-spark · -17.8 rating · Closed weightsgpt-5.5-high · -20.1 rating · Closed weightsglm-5.3-max · -20.5 rating · Open weightsclaude-sonnet-5.5-xhigh · -20.5 rating · Closed weightsglm-5.3-flash · -20.6 rating · Open weightsqwen3.5-max-preview · -21.3 rating · Closed weightsgpt-5.5 · -24.9 rating · Closed weightsgemini-2.5-pro · -34.0 rating · Closed weightsgpt-5.4 · -39.6 rating · Closed weightsdeepseek-v4-pro · -40.6 rating · Open weightsgrok-4.5 · -43.6 rating · Closed weightsqwen3.8-27b · -50.6 rating · Open weightshy3 · -52.8 rating · Open weightsmistral-large-4 · -64.3 rating · Closed weightsdeepseek-v3.2 · -67.1 rating · Open weightsgemini-2.5-flash · -74.8 rating · Closed weightsgrok-4.3 · -94.8 rating · Closed weightsdeepseek-r1 · -119.0 rating · Open weightsgpt-oss-120b · -126.2 rating · Open weightsgemma-3-27b-it · -133.9 rating · Open weightsdeepseek-v3 · -159.2 rating · Open weightsgpt-4o-2024-05-13 · -191.2 rating · Closed weightsclaude-3-7-sonnet-20250219 · -192.3 rating · Closed weightsclaude-3-5-sonnet-20241022 · -193.9 rating · Closed weightsllama-3.3-70b-instruct · -217.4 rating · Open weightsmistral-large-2411 · -226.4 rating · Weights unknownHorizontal placement starts from archive debut, then separates overlapping models.
8 Oct 2026
GoogleAnthropicXiaomiMetaAlibabamoonshotOpenAIzaiDeepSeekxAITencentMistral
Board published · 8 Oct 2026