Mozilla releases llamafile for portable local model inference

Release and source context

Mozilla’s innovation group and Justine Tunney released llamafile, combining model inference code and weights in portable executables. The project used llama.cpp and Cosmopolitan to simplify local model installation across supported operating systems and architectures. [1]

Mozilla funded the collaboration with Justine Tunney. The release packaged model weights and inference software using llama.cpp and Cosmopolitan; it did not introduce a newly trained model. [1]