Model rankings
| # | Model | Top images | Score | Evidence |
|---|---|---|---|---|
| 1 | GPT-6.1 Sol | 12371094–1380 | 20+ votes | |
| 2 | Kimi K3.1 Pro | 11951022–1369 | 20+ votes | |
| 3 | GPT-6.1 Terra | 1178649–1706 | Under 10 votes | |
| 4 | GPT-6 Astra | 1140996–1284 | 20+ votes | |
| 5 | Opus 5.5 | 1108969–1246 | 20+ votes | |
| 6 | Sonnet 5.5 | 1068938–1197 | 20+ votes | |
How scores work
Every model starts at 1000. Each vote is a small match between two models: the winner gains points and the loser drops. Beating a high-scoring model earns more than beating a low-scoring one, and a tie counts as half a win for each. The small range under each score shows how sure the score is; it narrows as more votes come in.
Technical details
Scores use a regularized Bradley–Terry model on an Elo scale, starting at 1000. Opponents and outcomes determine the rating; a tie expresses equal preference. Overall ratings use all published image comparisons, without subject or style weighting. Subject and style views fit their own comparison subsets.
Displayed ranges show approximate 95% uncertainty from the fitted model. Sparse estimates stay closer to the starting score. The info icon provides evidence details for estimates with limited votes or opponent coverage.
Learn more: Elo rating system ↗ · Bradley–Terry model ↗
Head-to-head win rates
Models run newest to oldest. Each cell shows how often the model on the left beat the model on top. Ties count as half a win. Empty cells haven’t been compared yet.
| Model | GPT-6.1 Terra | Kimi K3.1 Pro | Muse Wonder Full | Haiku 5.5 | GPT-6.1 Sol | Sonnet 5.5 | Opus 5.5 | GPT-6 Luna | GPT-6 Astra | Fable 5.1 | GPT-5.6 Terra |
|---|---|---|---|---|---|---|---|---|---|---|---|
| GPT-6.1 Terra | 100 | 100 | |||||||||
| Kimi K3.1 Pro | 100 | 100 | 50 | 50 | 33 | 100 | 50 | 100 | 100 | ||
| Muse Wonder Full | 0 | 40 | 0 | 0 | 0 | 50 | 50 | 0 | 100 | ||
| Haiku 5.5 | 0 | 0 | 60 | 0 | 20 | 0 | 80 | 0 | 40 | 50 | |
| GPT-6.1 Sol | 50 | 100 | 100 | 80 | 83 | 75 | 38 | 80 | 100 | ||
| Sonnet 5.5 | 50 | 100 | 80 | 20 | 50 | 90 | 40 | 50 | 67 | ||
| Opus 5.5 | 67 | 100 | 100 | 17 | 50 | 50 | 0 | 100 | 100 | ||
| GPT-6 Luna | 0 | 0 | 50 | 20 | 25 | 10 | 50 | 25 | 0 | 50 | |
| GPT-6 Astra | 50 | 50 | 100 | 63 | 60 | 100 | 75 | 50 | 67 | ||
| Fable 5.1 | 0 | 100 | 60 | 20 | 50 | 0 | 100 | 50 | 100 | ||
| GPT-5.6 Terra | 0 | 0 | 50 | 0 | 33 | 0 | 50 | 33 | 0 |
Hover or tap a cell for details.
100+ comparisons · Updated 4 h ago.