Video-Centered Research - Claude Code Successful Average: leaderboard
Video-Centered DeepResearch: 50 tasks with a frozen video, query, report and tool contract; Claude Code native harness, 64K tokens and 500 iterations, up to three complete attempts and 10 hours per attempt. All-task average assigns zero to failed reports; quality dimensions and successful average use each model’s successful subset. GPT-5.6 Sol provides semantic and visual judging alongside deterministic browser checks. Generation effort is undisclosed. Score (%).
Metric: Score (%). Source: arxiv.org. Saturation forecast: Around August 2027. 12 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5 | 66 | #13 |
| 2 | Claude Opus 4.8 | 62 | #37 |
| 3 | GLM-5.3 | 62 | #38 |
| 4 | Claude Opus 4.7 | 60.9 | #52 |
| 5 | Qwen 3.8 Max | 60.1 | #32 |
| 6 | Kimi K3 | 55.1 | #33 |
| 7 | GPT-5.6 Sol | 54.9 | #19 |
| 8 | Gemini 3.8 Flash | 52.6 | #23 |
| 9 | GPT-5.5 | 50.4 | #30 |
| 10 | Qwen 3.7 Max | 45.4 | #79 |
| 11 | Qwen 3.5 Omni Plus | 45.4 | #104 |
No result here: #5 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=video-centered-research-claude-code-successful-average · How It Works · Data refreshed daily, snapshot 2026-10-25.