SuperCLUE General (March 2026) - Precise Instruction Following: leaderboard

Metric: Score. Source: www.superclueai.com. 24 models tracked.

Top models

#ModelScore
1Gemini 3.1 Pro (Preview) (High)56.76
2Claude Opus 4.6 (Max)47.57
3GPT-5.4 (xHigh)44.32
4DeepSeek V4 Pro (Max)37.84
5Gemini 3 Flash (Preview) (High)35.68
6DeepSeek V4 Flash (Max)32.43
7Grok 4.20 Beta (0309) (Reasoning)32.43
8DeepSeek V3.2 (Thinking)25.95
9GLM-524.86
10GPT-OSS-120B (High)20.54
11Qwen 3.5 397B A17B (Thinking)19.46
12MiniMax-M2.717.3
13Kimi K2.5 (Thinking)16.22
14MiMo-V2-Pro16.22
15Tencent HY 2.0 Think14.05

Interactive version: theaggregate.ai/benchmark?slug=superclue-general-march-2026-precise-instruction-following · How It Works · Data refreshed daily, snapshot 2026-09-19.