Video-Centered Research - Codex All-task Average: leaderboard
Video-Centered DeepResearch: 50 tasks with a frozen video, query, report and tool contract; Codex 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 October 2027. 3 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.6 Sol | 53.9 | #19 |
| 2 | GPT-5.5 | 46 | #30 |
| 3 | Qwen 3.7 Max | 45.3 | #79 |
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-codex-all-task-average · How It Works · Data refreshed daily, snapshot 2026-10-25.