IndustryBench (Thinking): leaderboard

IndustryBench (2026) answered with thinking mode enabled: 2,049 closed-book Chinese industrial procurement questions grounded in national standards, scored 0 to 3 by a Qwen3-Max judge against reference answers, with any answer that contradicts 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, thinking mode enabled, only the final answer judged, over the 2,049 Chinese industrial procurement QA items of IndustryBench grounded in Chinese national standards (GB/T); higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 13 models tracked. Updated 7 Oct 2026.

Top models

#ModelSettingScoreOverall rank
1Claude Opus 4.6Thinking2.03#114
2GPT-5.4Thinking1.98#138
3Gemini 3.1 Pro (Preview)1.97#102
4Qwen 3.6 PlusThinking1.89#245
5Qwen 3.5 397B A17B1.81#270
6Qwen 3.5 PlusThinking1.79#250
7Qwen 3 MaxThinking1.75#420
8GLM-51.72#276
9Qwen 3.5 122B A10B1.71#350
10Kimi K2.51.68#272
11Qwen 3.5 27B1.65#380
12Qwen 3.5 35B A3B1.64#494
13MiniMax-M2.51.42#600

No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=industrybench-thinking · How It Works · Data refreshed daily, snapshot 2026-11-22.