IndustryBench - Safety and Compliance: leaderboard

IndustryBench (2026) restricted to its 116 Safety and Compliance questions (safety standards, risk mitigation and regulatory compliance), Chinese industrial procurement QA answered closed-book in default non-reasoning mode and scored 0 to 3 by a Qwen3-Max judge, with answers that violate a safety clause set to 0. Mean score out of 3; higher is better.

Metric: Final (SV) score (out of 3): mean 0-3 rubric score from a Qwen3-Max judge against the reference answer, set to 0 for any answer with a safety violation against the source standard; closed-book, zero-shot, provider-default decoding, over the 116 items of the 2,049 Chinese industrial procurement QA items of IndustryBench grounded in Chinese national standards (GB/T) in the Safety and Compliance capability dimension (safety standards, risk mitigation and regulatory compliance; default non-reasoning runs); higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 17 models tracked. Updated 7 Oct 2026.

Top models

#ModelSettingScoreOverall rank
1Qwen 3.6 PlusNon-reasoning2.4#238
2GPT-5.4Non-reasoning2.37#135
3Qwen 3.5 397B A17BNon-reasoning2.37#264
4GPT-5.2Non-reasoning2.34#185
5Qwen 3.5 PlusNon-reasoning2.21#242
6Qwen 3.5 27BNon-reasoning2.15#375
7Gemini 3.1 Pro (Preview)2.14#99
8Claude Opus 4.62.08#113
9Qwen 3.5 122B A10BNon-reasoning2.07#346
10Qwen 3.5 35B A3BNon-reasoning2.03#492
11GLM-5Non-reasoning2.02#269
12Kimi K2.5Non-reasoning1.94#266
13Qwen 3 MaxNon-reasoning1.92#415
14Claude Sonnet 4.61.92#173
15MiniMax-M2.51.76#600

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=industrybench-safety-and-compliance · How It Works · Data refreshed daily, snapshot 2026-11-21.