ReLE - Language and Instruction Following: leaderboard
ReLE Chinese LLM benchmark (NoneLinear, formerly CLiB), a continuously updated Chinese-language evaluation run through each model's API. Language and instruction-following domain: idioms, sentiment, entailment, classification, information extraction, reading comprehension, pronoun resolution, poetry matching, Chinese instruction following and character glyphs. Accuracy 0-100, higher is better.
Metric: Accuracy (%). Source: nonelinear.com. Saturation forecast: Around October 2027. 148 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 4 Argon | 81.6 | #9 |
| 2 | Hy4 preview | 78.8 | #45 |
| 3 | DeepSeek V4 Pro | 77 | #112 |
| 4 | GLM-5.3 Flash | 77 | #77 |
| 5 | Seed 2.0 Pro | 76 | #122 |
| 6 | GPT-5 | 75.6 | #105 |
| 7 | GPT-6 Astra | 75.3 | #6 |
| 8 | GLM-5.3 | 75 | #44 |
| 9 | DeepSeek V3.2 (Thinking) | 74.7 | #210 (DeepSeek V3.2) |
| 10 | Claude Opus 5.5 | 74.7 | #3 |
| 11 | GPT-6.1 Sol | 74.5 | #7 |
| 12 | GPT-6 Luna | 74.3 | #61 |
| 13 | DeepSeek V4.1 Flash | 74.1 | #37 |
| 14 | Gemini 3.8 Flash | 73.5 | #30 |
| 15 | Kimi K2.7 Code | 73 | #95 |
No result here: #10 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=rele-language-and-instruction-following · How It Works · Data refreshed daily, snapshot 2026-11-05.