Colon-Bench - VQA (Prompted): leaderboard
Metric: Accuracy (%) on Colon-Bench's 1,485 prompted five-option video questions (lesion identification, clinical characteristics, temporal reasoning) over 499 clips shown with lesion box and mask overlays; unanswered questions count as wrong; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 17 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro | 78.6 | #77 |
| 2 | Gemini 3 Flash | 76.6 | #93 |
| 3 | Gemini 3.1 Flash Lite | 69.2 | #174 |
| 4 | Seed-1.6 | 62.9 | #257 |
| 5 | GPT-5.2 | 62 | #105 |
| 6 | GPT-5.4 | 59.4 | #76 |
| 7 | GLM-4.6V | 55.7 | #309 |
| 8 | Qwen 3.5 397B A17B | 49 | #141 |
| 9 | Gemini 2.5 Flash Lite | 47.8 | #413 |
| 10 | Qwen 3 VL 235B A22B | 46.7 | #228 |
| 11 | Molmo2-8B | 46.1 | #522 |
| 12 | Qwen 3.5 Plus | 44.3 | #123 |
| 13 | Seed 1.6 Flash | 38.1 | #427 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=colon-bench-vqa-prompted · How It Works · Data refreshed daily, snapshot 2026-10-11.