ALL Bench Multimodal: leaderboard
Composite multimodal leaderboard aggregating model results across VLM, image generation, video generation, and agent-style multimodal evaluations.
Metric: Average Numeric VLM Score (%). Source: github.com. Source: huggingface.co. Saturation forecast: Around January 2027. 16 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.2 | 86.7 | #110 |
| 2 | Claude Opus 4.6 | 85.1 | #63 |
| 3 | GPT-5 | 84.2 | #98 |
| 4 | Gemini 3 Flash | 83.8 | #99 |
| 5 | Gemini 3 Pro | 83.75 | #80 |
| 6 | Gemini 3.1 Pro (Preview) | 82 | #58 |
| 7 | InternVL3-78B | 79.74 | #352 |
| 8 | Qwen 3.5 9B | 71.32 | #379 |
| 9 | Qwen 3.5 4B | 68.72 | #475 |
| 10 | GPT-5 Nano | 55.24 | #425 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=all-bench-multimodal · How It Works · Data refreshed daily, snapshot 2026-10-20.