MMLongBench-Doc (multimodal long-document benchmark) — Distinct from M-LongDoc (diff benchmark family) despite domain similarity.
| # | Model | Vendor | Best score | Runs | Last seen |
|---|---|---|---|---|---|
| 1 | GPT-4o | OpenAI | 42.8 | 2 | 2026-06-05 |
| 2 | Kimi VL A3B Thinking | Moonshot | 42.1 | 2 | 2026-06-05 |
| 3 | Qwen2.5 VL 72B | Alibaba | 38.8 | 1 | 2026-06-05 |
| 4 | Kimi VL A3B Instruct | Moonshot | 35.1 | 2 | 2026-06-12 |
| 5 | GPT-4o mini | OpenAI | 29 | 1 | 2026-06-05 |
| 6 | Gemma 3 12B | 21.3 | 2 | 2026-06-05 |
Best tracked score per model (default configuration; source-attributed and verification-tiered). Open a model for its full benchmark surface, provenance and pricing.