HELM NaturalQuestions (Closed): leaderboard
Closed-book question answering from Google search queries without passage access, testing parametric knowledge recall. Part of Stanford HELM.
Metric: F1 (%). Source: crfm.stanford.edu. Saturation forecast: Estimated already saturated. 91 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude 3.5 Sonnet (20240620) | 50.16 | #299 |
| 2 | GPT-4o (2024-05-13) | 50.13 | #325 |
| 3 | GPT-4o (2024-08-06) | 49.58 | #336 |
| 4 | GPT-4 Turbo | 48.21 | #404 |
| 5 | Mixtral 8x22B | 47.76 | #766 |
| 6 | Llama 3 70B | 47.53 | #631 |
| 7 | Claude 3.5 Sonnet (20241022) | 46.7 | #291 |
| 8 | DeepSeek V3 | 46.7 | #321 |
| 9 | Llama 2 70B | 45.96 | #1201 |
| 10 | GPT-4 (0613) | 45.69 | #491 |
| 11 | Llama 3.2 90B Vision Instruct | 45.68 | #622 |
| 12 | Llama 3.1 405B Instruct | 45.6 | #457 |
| 13 | Gemini 1.5 Pro (002) | 45.55 | #349 |
| 14 | Mistral Large 2 (Jul) | 45.27 | #484 |
| 15 | Llama 3.1 70B Instruct | 45.17 | #561 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=helm-naturalquestions-closed · How It Works · Data refreshed daily, snapshot 2026-10-20.