| # | Model | Vendor | Best score | Runs | Last seen |
|---|---|---|---|---|---|
| 1 | InternVL3-2B | OpenGVLab | 90.1 | 1 | 2026-08-09 |
| 2 | LFM2-VL-3B (3.1B) | Liquid AI | 89.2 | 1 | 2026-08-12 |
| 3 | LFM2-VL-3B | Liquid AI | 89 | 1 | 2026-08-09 |
| 4 | LFM2.5-VL-3B | Liquid AI | 88.7 | 3 | 2026-08-23 |
| 5 | InternVL3_5-2B | OpenGVLab | 87.2 | 1 | 2026-08-09 |
| 6 | LFM2.5-VL-450M-Extract | Liquid AI | 86.9 | 1 | 2026-08-09 |
| 7 | Qwen2.5 VL 3B | Alibaba | 86.2 | 1 | 2026-08-09 |
| 8 | Phi 3.5 Vision Instruct | Microsoft | 86.1 | 1 | 2026-08-23 |
| 9 | Phi 4 Multimodal Instruct | Microsoft | 85.6 | 1 | 2026-08-23 |
| 10 | Gemma 4 E2B | 84 | 1 | 2026-08-12 | |
| 11 | LFM2-VL-450M | Liquid AI | 83.8 | 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.