Georgi Gerganov, creator of the widely used llama.cpp and whisper.cpp projects, today announced the formation of ggml.ai, a company dedicated to developing the ggml tensor library. The library, already powering local inference of large language models on commodity hardware, now has formal financial backing: Nat Friedman and Daniel Gross provided pre-seed funding.

ggml is designed as a minimal, open-core tensor library with integer quantization support, broad hardware compatibility, and zero memory allocations during runtime. It is used by llama.cpp and whisper.cpp for local inference on commodity hardware.

The Hacker News community reacted with a mix of excitement and skepticism. Commenters celebrated the potential for private, fine-tuned local AI assistants, with one user calling it “brilliant” to have financial support. Others voiced concerns that venture funding could eventually push the project toward a less open model, noting that the website already states “in the future we may choose to develop extensions that are licensed for commercial use.”

For now, the project remains fully MIT-licensed, and contributors are encouraged to “try crazy ideas, build wild demos, and push the edge of what’s possible.”