CapArena-Auto vs GPT-4o: leaderboard
CapArena-Auto vs GPT-4o measures pairwise caption wins and losses against the GPT-4o baseline.
Metric: CapArena-Auto Score (vs GPT-4o). Source: huggingface.co. Saturation forecast: Estimated already saturated. 22 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 1.5 Pro (002) | 29 | #349 |
| 2 | GPT-4o (2024-08-06) | 0 | #336 |
| 3 | Qwen 2.5 VL 72B Instruct | -1 | #377 |
| 4 | Gemini 2.0 Flash (Preview) | -2 | #370 |
| 5 | Claude 3.5 Sonnet (20240620) | -14 | #299 |
| 6 | GPT-4o Mini (2024-07-18) | -36 | #558 |
| 7 | Llama 3.2 90B Vision Instruct | -72 | #622 |
| 8 | Qwen 2 VL 7B Instruct | -78 | #826 |
| 9 | MiniCPM-V-2.6 | -80 | #836 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=caparena-auto-vs-gpt-4o · How It Works · Data refreshed daily, snapshot 2026-10-20.