By Elo rating, ChatGPT 4o is expected to win 53% of comparisons against Seedream 4.5.
Both models received the same prompts. Wins and losses count only images judged together in the same call, the same comparisons the leaderboard Elo is fitted from.
| Category | ChatGPT 4o | Seedream 4.5 | Direct record |
|---|---|---|---|
| Anime & Cartoon Style | 746 | 964 | — |
| Architecture & Interiors | 910 | 884 | — |
| Complex Scenes | 1054 | 1081 | — |
| Ghibli style | 813 | 1020 | — |
| Graphic Design | 1099 | 924 | — |
| Hands & Anatomy | 933 | 961 | — |
| Photorealistic People & Portraits | 1078 | 935 | — |
| Surreal & Creative Prompts | 923 | 877 | — |
| Text in Images | 1003 | 943 | — |
| Ultra Hard | 1004 | 977 | — |
| Criterion | ChatGPT 4o | Seedream 4.5 |
|---|---|---|
| Prompt adherence | 1020 | 961 |
| Technical quality | 964 | 919 |
| Artistic merit | 901 | 991 |
| Detail execution | 924 | 955 |
| Realism & coherence | 995 | 952 |