GitHub radar
Hemmingway-1: Free local model for generating human-like text
Altworld’s open 27B chat model: Qwen3.8-27B without the small talk. Just the message. Your email. Your response to a coworker. That thing you’ve been meaning to write.
Hemmingway-1 is a 27B chat model fine tuned by Altworld (an indie studio) on top of Qwen3.8-27B. This model is optimized for common tasks such as email writing or quick responses to colleagues. According to the creators, this model is more human sounding and performs better than other larger models on their internal tests, however those results are not independently verified and should be taken with a grain of salt. With a large context size of 262,144 tokens, this model is able to process large amounts of text at once, such as entire chat logs. The weights are available for public download and can be used offline for non-commercial purposes. Prebuilt android and mac applications are also available.
Why a vibe-coder should care
If you want to write emails, messages, or quick responses this model is perfect for that. However, if you need anything more complex than that, I wouldn’t rely on this model too much. This is a small 27B model, and it’s not meant to replace a model like Claude or GPT. However, it does provide free non-commercial use and allows you to run the model offline on your own computer.
How to install
Copy this and send it to your agent — Claude Code, Codex, any of them:
Set up the Hemmingway-1 model from Altworld locally for me: https://huggingface.co/Altworld/Hemmingway-1 — pick a version that fits my hardware, run it through Ollama or LM Studio, and have it draft a short support message about a broken door lock.
You need a Mac with 32 GB of memory or more, or a GPU with 24 GB of VRAM — the 4-bit version takes about 16 GB, the original about 54 GB.
Open on Hugging Face▌ More finds