Hemingway-bench
Chat & Writing · 2026-02-04
Elo rating on real-world writing tasks graded by expert writers.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Fable 5 | 1118 |
| Gemini 3.6 Flash | 1076 |
| Gemini 3.1 Pro Preview | 1069 |
| Kimi K3 | 1065 |
| Gemini 3.5 Flash | 1064 |
| GPT-5.6 Sol | 1060 |
| Claude Opus 5 | 1057 |
| Gemini 3 Pro Preview | 1053 |
| Gemini 3 Flash Preview | 1051 |
| Opus 4.6 | 1043 |
| Opus 4.7 | 1039 |
| Opus 4.8 | 1038 |
| GPT-5.5 | 1036 |
| GLM-5.2 | 1032 |
| Kimi K2.6 | 1026 |
| Opus 4.5 | 1014 |
| DeepSeek-V4-Pro | 998 |
| Sonnet 4.6 | 996 |
| Kimi K2.5 | 991 |
| GPT-5.2 Instant | 989 |
| Qwen3.5 Plus (2026-02-15) | 989 |
| Gemini 3.5 Flash-Lite | 984 |
| GPT-5.4 | 980 |
| Muse Spark 1.1 | 978 |
| DeepSeek-V4-Flash | 976 |
| Grok 4.5 | 973 |
| GPT-5.2 | 954 |
| Qwen3 Max (rolling alias) | 948 |
| Grok 4.1 Fast | 918 |
| Kimi K2 Instruct | 885 |
| Llama 4 Maverick | 806 |
| Nova 2 Pro (Preview) | 754 |