HELM Classic - NaturalQuestions Closed Book — leaderboard

Metric: F1 (%). Source: crfm.stanford.edu. 67 models tracked.

Top models

#ModelScore
1Llama 2 70B45.84
2LLaMA-65B43.14
3LLaMA-30B40.83
4text-davinci-00340.64
5falcon-40B39.22
6GPT-3.5 Turbo (0301)39.01
7text-davinci-00238.31
8Llama 2 13B37.59
9Mistral-7B-v0.136.48
10GPT-3.5 Turbo (0613)34.76
11mpt-30B34.69
12LLaMA-13B34.58
13Llama 2 7B33.66
14davinci32.86
15LLaMA-7B29.75

Interactive version: theaggregate.ai/benchmark?slug=helm-classic-naturalquestions-closed-book · How the rankings work · Data refreshed daily, snapshot 2026-07-22.