HELM TORR - Tablebench Fact Checking - Robustness: leaderboard

Metric: Robustness (0-100). Source: crfm.stanford.edu. 14 models tracked.

Top models

#ModelScore
1Claude 3.5 Sonnet (20241022)60.22
2DeepSeek V359.75
3GPT-4o (2024-11-20)59.08
4Gemini 1.5 Pro (002)49.78
5Gemini 1.5 Flash (002)43.48
6Qwen 2 72B Instruct39.07
7Claude 3.5 Haiku (20241022)32.88
8GPT-4o Mini (2024-07-18)32
9Llama 3.1 405B Instruct21.84
10Llama 3.1 70B Instruct19.54
11Mistral 7B Instruct (v0.3)18.92
12Llama 3.1 8B Instruct9.85

Interactive version: theaggregate.ai/benchmark?slug=helm-torr-tablebench-fact-checking-robustness · How It Works · Data refreshed daily, snapshot 2026-09-19.