HELM: leaderboard
HELM: Evaluates broad language-model knowledge, reasoning, commonsense, instruction following, or exam-style accuracy.
Metric: Mean win rate (imported). Source: crfm.stanford.edu. Source: benchmarklist.com. Saturation forecast: Estimated already saturated. 66 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Llama 2 70B | 94.35 | #1201 |
| 2 | LLaMA-65B | 90.83 | #1303 |
| 3 | text-davinci-002 | 90.5 | #1225 |
| 4 | Mistral-7B-v0.1 | 88.4 | #1326 |
| 5 | text-davinci-003 | 87.16 | #1146 |
| 6 | Llama 2 13B | 82.3 | #1436 |
| 7 | GPT-3.5 Turbo (0613) | 78.3 | #1118 |
| 8 | LLaMA-30B | 78.13 | #1423 |
| 9 | Palmyra X (43B) | 73.25 | #1317 |
| 10 | falcon-40B | 72.94 | #1442 |
| 11 | falcon-40B Instruct | 72.66 | #1458 |
| 12 | mpt-30B Instruct | 71.64 | #1470 |
| 13 | mpt-30B | 71.45 | #1476 |
| 14 | vicuna-13B-v1.3 | 70.63 | #1447 |
| 15 | vicuna-7B-v1.3 | 62.53 | #1507 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=helm · How It Works · Data refreshed daily, snapshot 2026-10-20.