Description
Running an LLM on an embedded board is assumed to need gigabytes of RAM and Python. PicoLM, in pure C, runs a 1-billion-parameter LLM on a $10 board with 256MB of RAM with zero dependencies, one binary and no cloud.
It supports ARM and RISC-V, reads GGUF models, and serves as the local brain for the ultralight PicoClaw assistant.
Tiny footprint:1B model in 256MB.
Pure C:One dependency-free binary.
Architectures:ARM and RISC-V.
Offline:No cloud.
It supports ARM and RISC-V, reads GGUF models, and serves as the local brain for the ultralight PicoClaw assistant.
Features
Tiny footprint:1B model in 256MB.
Pure C:One dependency-free binary.
Architectures:ARM and RISC-V.
Offline:No cloud.
