SFI-Bench: leaderboard
Metric: Macro-average accuracy (%) over the six tasks (global and conditional counting, cross-view path reasoning, layout inference, functional association, operation planning, troubleshooting), four-option multiple choice on egocentric indoor videos (ARKitScenes and ScanNet++ scans), zero-shot with the same prompt templates, models at their default configurations; the first four tasks are answered directly and, for the last two, models without a search tool also answer offline; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 17 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Qwen 3 VL 235B A22B Instruct | 60.7 |
| 2 | Qwen 3 VL 32B Instruct | 59 |
| 3 | Qwen 3 VL 235B A22B (Thinking) | 57.9 |
| 4 | Qwen 3 VL 32B (Thinking) | 55.9 |
| 5 | Qwen 3 VL 8B Instruct | 53.3 |
| 6 | Qwen 3 VL 30B A3B Instruct | 52.7 |
| 7 | Qwen 3 VL 30B A3B (Thinking) | 52.1 |
| 8 | Qwen 3 VL 8B (Thinking) | 51.4 |
Interactive version: theaggregate.ai/benchmark?slug=sfi-bench · How It Works · Data refreshed daily, snapshot 2026-10-07.