b8073
Massive open-source update brings powerful new multilingual models to local AI.
Deep Dive
The popular open-source project llama.cpp has released version b8073, adding official support for Cohere's new Tiny Aya models. This update includes critical changes to the model hash, vocabulary, and tokenizer to handle the new architecture. The release provides pre-built binaries for macOS, iOS, Windows, and Linux, including CUDA, Vulkan, and HIP backends. This integration allows developers to immediately run the efficient, multilingual Tiny Aya models locally on their own hardware.
Why It Matters
This dramatically lowers the barrier for running state-of-the-art, efficient multilingual AI models on consumer devices.