Hemmingway-1, an Apache-2.0 27B creative-writing fine-tune (Qwen3.8-27B base, EQ-Bench 4 1330)[R]
Hemmingway-1 is a new Apache-2.0 licensed 27B creative-writing fine-tune, based on Qwen3.8-27B, developed by a small lab in Switzerland and South Africa. Specialized for short-form personal texts like stories, dialogue, and emails, it achieved an EQ-Bench 4 score of 1330. Internal benchmarks show it outperforms frontier models in communicationbench (1026) and human-likeness (1032). The model's focus is creative writing, with math, code, and factual recall remaining unchanged from the base model. It is available at hemmingway.io and its weights are on huggingface.co/Altworld/Hemmingway-1.
This model is notable for being the first open-sourced 27B fine-tune of Qwen3.8-27B specifically optimized for creative writing, unlike general-purpose models.
Time & source
Times shown in UTC
Display time zone: UTC
Local time zone unavailable; showing UTC.
IngestedOffset at this time: UTC+0Sep 21, 2026, 07:01 UTC
- Ingested
- Sep 21, 2026, 07:01
- Source type
- Dev community
Full text isn't available here.
Read at source →