Video-HolmesV2 - Reasoning: leaderboard
Video-HolmesV2 reasoning average over intra-clip, inter-clip, counting and theme questions that need cross-modal audio-visual evidence chains. Higher is better. Source: arxiv.org/abs/2609.17248.
Metric: Average (%). Source: arxiv.org. Saturation forecast: Around February 2027. 14 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Pro | 73.1 | #159 |
| 2 | Gemini 2.5 Flash | 65.2 | #255 |
| 3 | Qwen 3 VL 235B A22B | 55.1 | #251 |
| 4 | InternVL3-8B | 48 | #624 |
| 5 | GPT-4o | 46.4 | #357 |
| 6 | Qwen2.5-Omni-7B | 40.6 | #633 |
| 7 | Qwen 2.5 VL 7B | 37.9 | #641 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=video-holmesv2-reasoning · How It Works · Data refreshed daily, snapshot 2026-11-05.