Best at what?

A composite of reasoning and knowledge evaluations. Higher scores are better; these are index points, not percent correct.

1. Claude Opus 5.5 (max with fallback)Intelligence Index
   58
2. Claude Sonnet 5.5 (max with fallback)Intelligence Index
   56
3. Claude Opus 5.5 (xhigh with fallback)Intelligence Index
   56
4. Claude Opus 5.5 (high with fallback)Intelligence Index
   54
5. Claude Fable 5.1 (max with fallback)Intelligence Index
   53

Effort settings and fallback runs affect results. An aggregate score can hide weaknesses on your particular task.

People compare anonymous model responses and choose the answer they prefer.

1. gemini-4-argon-highPreliminary · 4,932 votes
   1525±9
2. claude-opus-4-6-high77,636 votes
   1505±3
3. claude-fable-5-high38,387 votes
   1504±4
4. claude-opus-5.5-high4,552 votes
   1504±9
5. claude-opus-4-7-high64,946 votes
   1501±4

Preference is not a factual-accuracy score. Gemini 4 Argon has a restricted rollout; a leaderboard listing does not guarantee access. Uncertainty ranges overlap for many entries; their order is not proof of a meaningful difference.

Head-to-head preferences for generated web applications.

1. claude-opus-5.5-max2,062 votes
   1815+16/-16
2. gpt-6-astra-max6,123 votes
   1788+10/-10
3. claude-sonnet-5.5-xhigh1,531 votes
   1786+18/-18
4. gpt-6.1-sol-max1,620 votes
   1758+17/-17
5. claude-fable-5.1-max6,318 votes
   1749+10/-10

This measures web-building output, not every kind of coding, security or maintenance work. Uncertainty ranges overlap for many entries; their order is not proof of a meaningful difference.

Human preferences between images generated from prompts.

1. gpt-image-2.5-sunburstPreliminary · 10,884 votes
   1424±8
2. gpt-image-2.5-flarePreliminary · 10,058 votes
   1401±8
3. gpt-image-2 (medium)88,744 votes
   1383±4
4. mai-image-2.618,625 votes
   1335±6
5. grok-imagine-image-2.0 (canvas)Preliminary · 3,087 votes
   1335±12

A high rating does not guarantee accurate text, product details or permission to use an image commercially. Uncertainty ranges overlap for many entries; their order is not proof of a meaningful difference.

Human preferences between generated video clips.

1. gemini-omni-1.1-flash1,784 votes
   1516±15
2. gemini-omni-flash26,576 votes
   1513±9
3. flux-3-videoPreliminary · 1,302 votes
   1493±17
4. grok-imagine-video-1.5-agentPreliminary · 1,218 votes
   1492±18
5. dreamina-seedance-2.0-720p56,318 votes
   1479±8

Compare duration, resolution and audio separately. Early results can move substantially as votes accumulate. Uncertainty ranges overlap for many entries; their order is not proof of a meaningful difference.