WildVision-Bench: leaderboard
WildVision-Bench — real-world user visual queries drawn from WildVision-Arena testing vision-language chat in the wild; judged against a GPT-4o reference, reward score (-100 to 100).
Metric: Overall Reward. Source: huggingface.co. Saturation forecast: Estimated already saturated. 29 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude 3.7 Sonnet | 60.6 | #249 |
| 2 | GPT-4.1 (2025-04-14) | 58.7 | #250 |
| 3 | Claude 3.5 Sonnet (20241022) | 53.2 | #291 |
| 4 | InternVL3-38B | 45.2 | #405 |
| 5 | InternVL2.5-78B | 42.4 | #431 |
| 6 | Grok 2 (1212) | 40 | #496 |
| 7 | InternVL3-78B | 39.6 | #352 |
| 8 | InternVL3-8B | 36.5 | #620 |
| 9 | Gemini 2.0 Flash | 35.5 | #342 |
| 10 | Qwen 2.5 VL 7B | 35 | #643 |
| 11 | Gemini 2.5 Flash | 32.4 | #245 |
| 12 | Llama 4 Scout Instruct | 31.5 | #539 |
| 13 | Pixtral-12B | 11.8 | #808 |
| 14 | Llama 3.2 11B Instruct | 1.9 | #1126 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=wildvision-bench · How It Works · Data refreshed daily, snapshot 2026-10-20.