HELM Lite - MATH Level 1 - Geometry: leaderboard
HELM Lite v1.13.0 MATH chain-of-thought run on Level 1 (easiest) Geometry problems only: 38 test problems, up to 8-shot, scored by equivalence of the final answer to the reference. Percent, 0-100, higher is better; a small sample, not comparable with full-MATH Geometry scores.
Metric: Equivalent (CoT) (%). Source: crfm.stanford.edu. Saturation forecast: Estimated already saturated. 91 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 1.5 Flash (002) | 81.58 | #587 |
| 2 | DeepSeek V3 | 81.58 | #329 |
| 3 | Gemini 2.0 Flash (Preview) | 81.58 | #376 |
| 4 | Claude 3.5 Sonnet (20241022) | 78.95 | #302 |
| 5 | Gemini 1.5 Pro (002) | 78.95 | #368 |
| 6 | Qwen 2.5 72B Instruct | 76.32 | #469 |
| 7 | GPT-4o (2024-08-06) | 73.68 | #343 |
| 8 | Claude 3.5 Haiku (20241022) | 73.68 | #590 |
| 9 | Nova Pro | 73.68 | #594 |
| 10 | GPT-4 Turbo | 71.05 | #417 |
| 11 | Llama 3.1 405B Instruct | 71.05 | #470 |
| 12 | Mistral Large | 71.05 | #713 |
| 13 | Gemini 1.5 Pro (001) | 71.05 | #465 |
| 14 | GPT-4 Turbo (Preview) | 71.05 | #416 |
| 15 | Nova Micro | 71.05 | #891 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=helm-lite-math-level-1-geometry · How It Works · Data refreshed daily, snapshot 2026-11-05.