2026-09-20 · America/Los_Angeles · 社区动态 · #18

we made a 27b model for creative writing. performs as good as claude fable 5, at a 40x cheaper price, open weights.

hey! we're a small swiss and south african lab and we just put out hemmingway-1. it's a 27b model we trained specifically for essays, creative writing, dialogue, roleplay, and for everyday texting. frontier performance at a 40x cheaper price. we thought it might be useful for people here who like cheap models that still write well. it benchmarks at absolute frontier level. on eq-bench 4 it scores 1330, right behind claude fable 5, and beating many other frontier models including gpt-5.5 and opus 4.8. you can see our internal benchmarks on the model card: hemmingway.io/model/ the testers we…

热度 45.0 / 100；排名与评分保留该期记录。