Hemmingway-1 is a 27B parameter model released with open weights. It is free for non-commercial use under CC BY-NC 4.0. Commercial use requires a separate agreement. It outperforms larger models on everyday writing tasks. It ranks first in human-likeness tests. It excels in work, admin, and difficult requests. It struggles with hostile storytelling and long story turns. It minimizes commentary and options in responses. It placed third on the EQ-Bench 4 emotional intelligence benchmark. It performs well on StoryBench. It is built on Qwen3.8-27B. It supports a 262,144 token context window.
What changed
Most models return options and commentary instead of the requested text.
What it unlocks
Getting a direct, human-like message without extra commentary or options.
- 27B parameters
- 262,144 token context
- 50 points ahead of GPT-6 Astra on CommunicationBench
- 26 points clear on human-likeness
What you need to act on it
- non-commercial use under CC BY-NC 4.0
- commercial use requires a separate agreement
Sources