OmniCoT-Real: leaderboard
OmniCoT-Real (2026), the real-world subset of the OmniCoT panoramic spatial reasoning benchmark: overall accuracy in percent on manually annotated questions about real indoor panoramas, with chain-of-thought prompting. Higher is better.
Metric: Overall accuracy (%), question-weighted over See, Locate and Move, on OmniCoT-Real (1,073 manually annotated questions on real-world indoor panoramas), chain-of-thought prompting. Source: arxiv.org. Saturation forecast: Around January 2028. 5 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Flash | 30.1 | #164 |
| 2 | GPT-5 | 29.45 | #167 |
| 3 | Qwen 2.5 VL 7B Instruct | 15.28 | #1833 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=omnicot-real · How It Works · Data refreshed daily, snapshot 2026-11-13.