MiniMax H3 Inference Engine h3.c Open Sourced with Mac and 3090 Support
QuixiAI · x · 2026-08-25
A new open-source inference engine, QuixiAI/h3.c, has been released for the MiniMax H3 model. The project is optimized for Mac devices and NVIDIA 3090 GPUs, supporting scaling from 1 to 8 GPUs. Key features include the addition of a user interface (UI) and GGUF support, allowing a single RTX 3090 to run the model at BF16 precision, aiming to lower the barrier for local deployment of the H3 model.
Related event: Open-source h3.c engine runs MiniMax H3 in under 10 minutes on 8x RTX 3090(5 posts)→
More from Infra
- Nvidia employee charged with smuggling advanced chips into China — iamKierraD · 2026-08-25
- Papers with Code Search Engine: Powered by Hugging Face Infrastructure — NielsRogge · 2026-08-25
- Llama.cpp adaptive speculation boosts inference speed by up to 50% — Dutchnamn · 2026-08-25
- Open Source RAG Stack: A Complete Architecture from Ingestion to Frontend — goyalshaliniuk · 2026-08-25
- UK data centres to emit more CO2 than ExxonMobil, analysis finds — nordicinst · 2026-08-25
- What is the next frontier for AI memory? — boneMechBoy69420 · 2026-08-25