Hemmingway-1 is a 27-billion-parameter, text-to-text causal language model built on Qwen3.8-27B and tuned for everyday English writing: messages, emails, workplace notes, difficult requests, and persuasive communication. It is published by HemmingwayAI; the model page lists The Commons Lab in Bülach, Switzerland as its maker. See the HemmingwayAI maintainer profile. The key reason to consider it is its stated preference for producing the requested message itself rather than a preamble, a menu of options, or commentary. Its advertised context window is 262,144 tokens, and the README provides Transformers and vLLM examples. The weights are available for non-commercial use under CC BY-NC 4.0, but commercial use requires a separate agreement. The model card does not disclose its training dataset, training procedure, hardware requirements, or inference speed, so its benchmark claims should be treated as the maintainer’s evidence rather than as a complete independent evaluation.
Best use cases
Drafting everyday messages and emails. Use it to write a note to a landlord about a broken boiler, request a refund, follow up on an unanswered email, or send a concise update to a colleague. The model was built for these routine communication tasks, and its CommunicationBench evaluation used 80 real requests in blind, pairwise comparisons. The maintainer reports that it ranked ahead of the tested alternatives on that benchmark. The benchmark is maintainer-created, so test it on your own prompts and tone requirements before adopting it.
