CG-Bench: leaderboard
Clue-grounded long-video question-answering benchmark evaluating MCQ accuracy, clue-grounding credibility metrics, and open-ended answer accuracy.
Metric: Long-Video MCQ Accuracy (%). Source: github.com. Source: cg-bench.github.io. Saturation forecast: Estimated already saturated. 22 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 2 VL 72B | 45.3 | #460 |
| 2 | GPT-4o (2024-08-06) | 44.9 | #336 |
| 3 | InternVL2.5-78B | 44.2 | #431 |
| 4 | Claude 3.5 Sonnet | 40.3 | #350 |
| 5 | Gemini 1.5 Pro | 37.8 | #445 |
| 6 | Gemini 1.5 Flash | 33.5 | #618 |
| 7 | GPT-4o Mini | 32.6 | #599 |
| 8 | MiniCPM-V-2.6 | 29.9 | #836 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=cg-bench · How It Works · Data refreshed daily, snapshot 2026-10-20.