ReLE - Language and Instruction Following: leaderboard

Metric: Accuracy (%). Source: nonelinear.com. 179 models tracked.

Top models

#ModelScore
1DeepSeek V4 Pro77
2GLM-5.3 Flash77
3Seed 2.0 Pro76
4GPT-575.6
5GPT-675.3
6GLM-5.375
7DeepSeek V3.2 (Thinking)74.7
8DeepSeek V4.1 Flash74.1
9Gemini 3.8 Flash73.5
10Kimi K2.7 Code73
11DeepSeek R1 052872.9
12Hunyuan-T1-2025071172.9
13GPT-5.6 Pro Sol72.8
14Qwen 3.8 Max72.6
15Kimi K372.5

Interactive version: theaggregate.ai/benchmark?slug=rele-language-and-instruction-following · How It Works · Data refreshed daily, snapshot 2026-09-19.