llama-modchip
github.com/synchopate/llama-modchip- Category
- AI Tools
- Rank
- No. 3227Tools index
- Pricing
- Open Source
- Type
- TOOL
- GitHub
- 430 stars
- Latest release
- v2.4.0
- Date
About
A set of llama.cpp and ggml kernel patches, quantization tools and runtime scripts for running large Mixture-of-Experts models on CPU only. It targets a 284B DeepSeek-V4-Flash checkpoint in under 59 GiB of RAM and 30B models on 16 to 64 GB machines, with no GPU. Running models this way matters for people who lack GPU hardware.
Tags
llminferencellama-cppcpumoequantization
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.