| # | Model | Vendor | Best score | Runs | Last seen |
|---|---|---|---|---|---|
| 1 | Qwen3 VL 4B Instruct | Alibaba | 92 | 1 | 2026-08-09 |
| 2 | InternVL3_5-4B | OpenGVLab | 90.2 | 1 | 2026-08-09 |
| 3 | Qwen3.5 2B | Alibaba | 89.7 | 2 | 2026-08-09 |
| 4 | InternVL3_5-2B | OpenGVLab | 87.7 | 2 | 2026-08-09 |
| 5 | Gemma 4 E2B | 84.4 | 2 | 2026-08-09 | |
| 6 | Qwen3.5 0.8B | Alibaba | 82.3 | 1 | 2026-08-09 |
| 7 | InternVL3_5-1B | OpenGVLab | 80.7 | 1 | 2026-08-09 |
| 8 | LFM2.5-VL-450M-Extract | Liquid AI | 73.4 | 1 | 2026-08-09 |
| 9 | LFM2.5-VL-1.6B | Liquid AI | 66 | 1 | 2026-08-09 |
| 10 | MiniCPM-V 4.6 | OpenBMB | 57.5 | 1 | 2026-08-09 |
Best tracked score per model (default configuration; source-attributed and verification-tiered). Open a model for its full benchmark surface, provenance and pricing.