Roboflow Playground Captioning — leaderboard
Metric: Arena Elo. Source: playground.roboflow.com. 48 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 2.5 Pro | 1287 |
| 2 | Qwen 3 VL 235B A22B Instruct | 1246 |
| 3 | Gemini 2.5 Flash | 1245 |
| 4 | Claude Opus 4.1 | 1236 |
| 5 | Gemma 3 12B | 1224 |
| 6 | Gemma 3 27B | 1223 |
| 7 | Qwen 3.5 9B | 1223 |
| 8 | Qwen 3.5 122B A10B | 1222 |
| 9 | Gemini 3 Flash | 1213 |
| 10 | GPT-5.2 | 1212 |
| 11 | Claude Sonnet 4.5 | 1212 |
| 12 | GPT-5.5 | 1212 |
| 13 | Claude Haiku 4.5 | 1212 |
| 14 | Claude Sonnet 5 | 1212 |
| 15 | GPT-5.6 Luna | 1212 |
Interactive version: theaggregate.ai/benchmark?slug=roboflow-playground-captioning · How It Works · Data refreshed daily, snapshot 2026-08-05.