Nano*
Decent AI for everyone.No GPU required.
Nano* is a family of small, fast runtimes that bring language, speech recognition and speech synthesis to ordinary computers: the laptops, desktops and servers people already own.
Three solutions
One idea, three products. Each runs on the CPU.
In development
NanoLLM
Language
Chat, tools and structured output with Gemma 4, quantized to 4 bits, served from one binary on the CPU.
Available
NanoASR
Speech recognition
An offline speech recognition server on sherpa-onnx: word timestamps, speakers, subtitles. One binary, on the CPU.
Preview
NanoTTS
Speech synthesis
A speech synthesis server for small models such as Pocket TTS and Supertonic 3, with an OpenAI-compatible API. CPU only.
No GPU required
Everything is tuned for the processor you already have: laptop, desktop or server.
Open source
MIT licensed. Read it, run it, change it.
All code lives on GitHub.