MathSpatial: leaderboard
MathSpatial — spatial-geometry math reasoning with diagrams; overall accuracy.
Metric: Overall Accuracy (%). Source: shuolucs.github.io. Saturation forecast: Around January 2027. 17 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5 | 58.5 | #98 |
| 2 | Gemini 2.5 Flash | 48.5 | #245 |
| 3 | Gemini 2.5 Pro | 44.9 | #152 |
| 4 | Claude Sonnet 4 | 26.7 | #199 |
| 5 | Llama 4 Maverick | 23.8 | #464 |
| 6 | GPT-4.1 | 22.6 | #248 |
| 7 | Claude 3.5 Sonnet | 22.3 | #350 |
| 8 | Claude 3.7 Sonnet | 21.5 | #249 |
| 9 | GLM-4.5V | 21 | #346 |
| 10 | GPT-4o | 19.6 | #343 |
| 11 | Qwen 2.5 VL 7B | 17.8 | #643 |
| 12 | InternVL3-8B | 17.4 | #620 |
| 13 | Llama 3 8B | 15 | #1242 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=mathspatial · How It Works · Data refreshed daily, snapshot 2026-10-20.