FreeToken
by FlashML-org
FreeToken is an engine for running very large open-weight AI language models on a home gaming PC or workstation by sharing the work between the graphics card, processor and system memory. It comes as a desktop app for Windows and Linux or as a command-line tool, and it offers OpenAI- and Anthropic-style APIs.
- Stars
- 16.3k
- Licence
- Apache-2.0
- Runs on
- Your computer, with a GPU
- Setup
- Moderate
- Language
- Python
- Age
- 105 days
What you can do with it
- Run large open-weight models such as DeepSeek-V4-Flash, Qwen or GLM on your own NVIDIA-equipped PC
- Chat with local models through the desktop app and tune engine settings
- Point coding agents such as Claude Code or Codex at a local OpenAI- or Anthropic-compatible API
Make money with it
Someone could offer private, on-premises AI setups for small firms that don't want to send data to cloud providers.
An idea, not financial advice. Check the licence before charging for anything built on it.
Ways a business could use it
- Run an in-house AI assistant or coding helper on your own hardware so client data stays private and you avoid per-token fees
- Power internal automations and agent workflows through a local API instead of paying for a hosted AI service
Can you use it commercially?
Free to use, change and sell, including in paid products. Keep the copyright notice.
Not legal advice. Read the project's licence file before building a product on it.
Getting it running
Download the desktop app for Windows or Linux, or install the command-line tool with uv or pip using "freetoken[accel]".
Follow the official instructions in the project README.
Things to check
- Needs a fairly powerful NVIDIA RTX 30, 40 or 50 series GPU and plenty of system memory
- Large models need big downloads and a lot of disk space
- Relatively new project, so expect rough edges and changing formats such as FTW checkpoints
Who's talking about it
Checked 03 Nov 2026. Summary drafted by AI from the project's README.
