External benchmark rankings
Pairwise human preference for multimodal responses.
Ranked among models offered by everyais · 3 models
| Rank | Model | Score |
|---|---|---|
| 1 | Claude Fable 5 Evaluation target / conditions: claude-fable-5 | 1,329 Elo |
| 2 | GLM-5.3 Flash Evaluation target / conditions: glm-5.3-flash | 1,296 Elo |
| 3 | Claude Opus 4.8 Evaluation target / conditions: claude-opus-4-8 | 1,289 Elo |
Each source, unit and index version has its own ranking. This is not the source’s full leaderboard. Equal scores share a rank.
Results use the source’s evaluation settings; they do not represent everyais default API settings.
Models without evaluation data are not ranked; missing data does not mean a low score.
Sync time is separate from the measurement date.