SVFSearch - Direct QA: leaderboard
Metric: Accuracy (%) on the 5,000-item SVFSearch test split (four-choice questions about a paused Chinese game video frame), given only the frame, question and options, answered directly without retrieval tools; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 12 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 77.5 |
| 2 | Claude Opus 4.7 | 69 |
| 3 | GPT-5.4 | 67.9 |
| 4 | Qwen 3.5 27B (Non-reasoning) | 66.4 |
| 5 | Qwen 3.5 27B (Thinking) | 64.8 |
| 6 | Qwen 3 VL 32B (Thinking) | 62.2 |
| 7 | Qwen 3.5 9B (Non-reasoning) | 59.9 |
| 8 | Qwen 3.5 9B (Thinking) | 58.3 |
| 9 | Qwen 3 VL 8B (Thinking) | 53.7 |
| 10 | Qwen 2.5 VL 7B | 49.8 |
Interactive version: theaggregate.ai/benchmark?slug=svfsearch-direct-qa · How It Works · Data refreshed daily, snapshot 2026-10-07.