HELM Classic - GSM8K: leaderboard
Stanford HELM Classic's run of GSM8K: grade-school mathematics word problems requiring multi-step arithmetic reasoning from natural-language questions. This is the GSM8K math scenario, not an aggregate of HELM's other tasks.
Metric: Exact Match (%). Source: crfm.stanford.edu. Saturation forecast: Estimated already saturated. 69 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Palmyra X (43B) | 63.3 | #1317 |
| 2 | text-davinci-003 | 50.6 | #1146 |
| 3 | Llama 2 70B | 48.4 | #1201 |
| 4 | GPT-3.5 Turbo (0613) | 46.9 | #1118 |
| 5 | LLaMA-65B | 46.6 | #1303 |
| 6 | text-davinci-002 | 41.53 | #1225 |
| 7 | Mistral-7B-v0.1 | 38.1 | #1326 |
| 8 | mpt-30B Instruct | 34.4 | #1470 |
| 9 | falcon-40B Instruct | 33.8 | #1458 |
| 10 | LLaMA-30B | 32 | #1423 |
| 11 | falcon-40B | 25 | #1442 |
| 12 | Llama 2 13B | 24.5 | #1436 |
| 13 | vicuna-13B-v1.3 | 22.6 | #1447 |
| 14 | mpt-30B | 16.4 | #1476 |
| 15 | LLaMA-13B | 15.4 | #1516 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=helm-classic-gsm8k · How It Works · Data refreshed daily, snapshot 2026-10-20.