OpenLMM Spatial - MMSI-Bench (Circular): leaderboard
MMSI-Bench from the OpenCompass OpenLMM Spatial Leaderboard (VLMEvalKit): multi-image spatial intelligence questions, scored with CircularEval (a question counts only if answered correctly under every rotation of its options). Accuracy in percent, higher is better.
Metric: CircularEval Accuracy (%). Source: huggingface.co. Saturation forecast: Around 2030. 58 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 | 21.8 | #105 |
| 2 | Gemini 2.5 Pro | 18.3 | #159 |
| 3 | InternVL3-38B | 17.7 | #413 |
| 4 | Qwen 2 VL 72B | 16.9 | #450 |
| 5 | InternVL3-78B | 16.6 | #361 |
| 6 | InternVL2.5-78B | 15.1 | #434 |
| 7 | GPT-5 Mini (2025-08-07) | 15 | #175 |
| 8 | InternVL3-8B | 14.7 | #624 |
| 9 | Claude 3.5 Sonnet (20241022) | 14.4 | #302 |
| 10 | InternVL3-14B | 14.4 | #506 |
| 11 | GPT-4.1 (2025-04-14) | 13.6 | #257 |
| 12 | Qwen 2.5 VL 7B | 13.4 | #641 |
| 13 | Claude 3.7 Sonnet | 13 | #259 |
| 14 | Gemini 2.0 Flash | 12.5 | #351 |
| 15 | Gemini 1.5 Pro (002) | 12.3 | #368 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=openlmm-spatial-mmsi-bench-circular · How It Works · Data refreshed daily, snapshot 2026-11-05.