HELM Lite: leaderboard
HELM Lite: 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. 77 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4o (2024-05-13) | 95.95 | #325 |
| 2 | Claude 3.5 Sonnet (20240620) | 91.22 | #299 |
| 3 | GPT-4 (0613) | 90.89 | #491 |
| 4 | GPT-4 Turbo | 89.84 | #404 |
| 5 | Llama 3.1 405B Instruct | 88.91 | #457 |
| 6 | Llama 3.2 90B Vision Instruct | 85.96 | #622 |
| 7 | Llama 3.1 70B Instruct | 84.8 | #561 |
| 8 | Llama 3 70B | 82.65 | #631 |
| 9 | Qwen 2 72B Instruct | 81.53 | #576 |
| 10 | Mistral Large 2 (Jul) | 78.77 | #484 |
| 11 | Gemini 1.5 Pro (001) | 78.18 | #444 |
| 12 | Qwen 2.5 72B Instruct | 78.06 | #454 |
| 13 | GPT-4o Mini (2024-07-18) | 75.68 | #558 |
| 14 | Mixtral 8x22B | 75.21 | #766 |
| 15 | Gemma 2 27B (IT) | 72.65 | #645 |
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 · How It Works · Data refreshed daily, snapshot 2026-10-20.