SFI-Bench - Global and Conditional Counting: leaderboard
Metric: Accuracy (%) on the global and conditional counting questions, four-option multiple choice on egocentric indoor videos (ARKitScenes and ScanNet++ scans), zero-shot with the same prompt templates, models at their default configurations; answered directly with no tools; higher is better. Source: arxiv.org. Saturation forecast: Around December 2027. 25 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 59.1 |
| 2 | GPT-5 | 58.4 |
| 3 | GPT-5.4 (High) | 58.4 |
| 4 | Gemini 3.1 Flash Lite | 55 |
| 5 | GPT-5.4 | 54.5 |
| 6 | Gemini 2.5 Pro | 54.4 |
| 7 | Qwen 3 VL 235B A22B (Thinking) | 53.8 |
| 8 | Qwen 3 VL 235B A22B Instruct | 52.3 |
| 9 | O4 Mini | 51 |
| 10 | Qwen 3 VL 32B Instruct | 50 |
| 11 | Qwen 3 VL 32B (Thinking) | 49.5 |
| 12 | Qwen 3 VL 8B (Thinking) | 42.6 |
| 13 | Qwen 3 VL 30B A3B Instruct | 42.1 |
| 14 | Gemini 2.5 Flash | 41.5 |
| 15 | Qwen 3 VL 8B Instruct | 41.5 |
Interactive version: theaggregate.ai/benchmark?slug=sfi-bench-global-and-conditional-counting · How It Works · Data refreshed daily, snapshot 2026-10-07.