LiveBench Instruction Following category score (0–100). Evaluates the ability to precisely follow complex multi-constraint instructions.
Source: LiveBench| Rank | Model | |
|---|---|---|
| #1 | Gemini 3.1 Pro | 79.1 |
| #2 | ChatGPT GPT-5.4 | 70.2 |
| #3 | Gemini 3.1 Flash Lite | 68.6 |
| #4 | Z.ai GLM-5.1 | 68.5 |
| #5 | Gemma 4 31B | 67.6 |
| #6 | ChatGPT GPT-5 Mini | 64.2 |
| #7 | Anthropic Claude Opus 4.6 | 63.3 |
| #8 | MiniMax MiniMax M2.7 | 61.1 |
| #9 | Qwen Qwen3.6 Plus | 58.3 |
| #10 | Moonshot AI Kimi K2.5 | 57.4 |
| #11 | MiniMax MiniMax M2.5 | 57.2 |
| #12 | Z.ai GLM-5 | 55.3 |
| #13 | ChatGPT GPT-5 Nano | 52.0 |
| #14 | Anthropic Claude Sonnet 4.6 | 51.6 |
| #15 | ChatGPT GPT OSS 120B | 50.3 |
| #16 | Xiaomi MiMo-V2-Pro | 43.2 |
| #17 | Gemini 2.5 Pro | 33.1 |
| #18 | Anthropic Claude Opus 4.5 | 28.9 |
| #19 | Gemini 2.5 Flash | 28.5 |
| #20 | Gemini 3 Flash | 28.3 |
| #21 | Anthropic Claude Sonnet 4.5 | 23.5 |
| #22 | Gemini 2.5 Flash Lite | 23.1 |
| #23 | DeepSeek DeepSeek V3.2 | 23.1 |
| #24 | ChatGPT GPT-5.4 Mini | 18.9 |
| #25 | Anthropic Claude Haiku 4.5 | 17.8 |
| #26 | ChatGPT GPT-5.4 Nano | 16.5 |