Hemingway-Bench: leaderboard
Surge AI compares model writing through pairwise preferences. The board reports native Elo ratings and confidence intervals, preserving each reasoning setting.
Metric: Elo (writing preference). Source: surgehq.ai. 47 models tracked.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | Claude Fable 5 | Adaptive Reasoning, High Effort | 1110 | #28 |
| 2 | Gemini 3.7 Flash | Thinking, Medium | 1105 | #38 |
| 3 | Claude Fable 5.1 | Adaptive Reasoning, High Effort | 1095 | #17 |
| 4 | Gemini 3.8 Flash | Thinking, Medium | 1093 | #39 |
| 5 | Gemini 3.6 Flash | Thinking, Medium | 1075 | #79 |
| 6 | Gemini 3.1 Pro | Thinking, High | 1073 | |
| 7 | Gemini 3.5 Flash | Thinking, Medium | 1071 | #107 |
| 8 | Kimi K3 | 1069 | #59 | |
| 9 | GPT-5.6 Sol | Thinking, Medium | 1057 | #33 |
| 10 | Gemini 3 Pro | Thinking, High | 1056 | #145 |
| 11 | GLM-5.3 | Thinking, Max | 1053 | #73 |
| 12 | Gemini 3 Flash | Thinking, High | 1052 | #174 |
| 13 | Claude Opus 5 | Adaptive Reasoning, High Effort | 1047 | #23 |
| 14 | DeepSeek V4 Pro | Thinking, High | 1046 | #198 |
| 15 | Claude Opus 4.6 | Thinking, High | 1045 | #114 |
No result here: #1 Claude Fable 5.2, #2 GPT-6.2 Astra, #3 Gemini 4 Carbon.
Interactive version: theaggregate.ai/benchmark?slug=hemingway-bench · How It Works · Data refreshed daily, snapshot 2026-11-17.