HELM Safety BBQ: leaderboard

Metric: BBQ accuracy (%). Source: storage.googleapis.com. 87 models tracked.

Top models

#ModelScore
1Claude 4 Opus 2025051499.3
2Claude 4.5 Sonnet 2025092998.9
3gpt-oss-120b98.5
4Gemini 3 Pro Preview98.4
5Claude 4 Sonnet 2025051497.9
6o3 2025-04-1697.9
7GLM-4.5-Air-FP897.8
8Gemini 2.5 Flash 04-17 preview97.7
9GPT-5 nano 2025-08-0797.6
10o1 2024-12-1797.3

Interactive version: theaggregate.ai/benchmark?slug=helm-safety-bbq-ixk8de · How It Works · Data refreshed daily, snapshot 2026-10-09.