OmniCoT - CoT F1: leaderboard
Reasoning quality on OmniCoT (2026), a panoramic spatial reasoning benchmark: the MME-CoT F1 times 100 of step precision and recall of reference solution steps, scored by an LLM judge. Higher is better.
Metric: Chain-of-thought F1 x100 (0-100): MME-CoT harmonic mean of step precision (faithfulness) and recall of reference solution steps, scored by an LLM judge on OmniCoT-B (synthetic 360-degree equirectangular panoramas with three anchor objects given by coordinates and orientation); answers are object identities, visibility, directions or the closest angle option, exact match after normalization. Source: arxiv.org. Saturation forecast: Around 2029. 21 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 | 41.51 | #167 |
| 2 | Qwen 3 VL 8B Instruct | 39.8 | #981 |
| 3 | Grok 4 | 36.4 | #307 |
| 4 | Qwen 2.5 VL 72B Instruct | 35.39 | #846 |
| 5 | Seed 1.8 | 35.13 | #242 |
| 6 | Gemini 3 Flash | 35.01 | #164 |
| 7 | Qwen 2.5 VL 32B Instruct | 33.17 | #1053 |
| 8 | ChatGPT 4o | 31.87 | #610 |
| 9 | GLM-4.5V | 31.12 | #744 |
| 10 | Qwen 2.5 VL 7B Instruct | 25.76 | #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-cot-f1 · How It Works · Data refreshed daily, snapshot 2026-11-13.