llama.cppProduct update
llama.cpp · v0.9.0
llama.cpp is a local large-language-model inference library with backends across CPU, CUDA, Metal and more.
v0.9.0 adds OCP FP8 E4M3 quantization and flash_attn_ext_rows so the unified KV cache stays efficient for multi-sequence inference. It also adds Clover, Limite, Metis-1.6 and Mistral Large 4 architectures, expands audio work such as Chatterbox TTS, and speeds Gemma 4 E2B/E4B prefill; ggml moves to v0.29.0.
This release adds OCP FP8 E4M3 quantization support, a new `flash_attn_ext_rows` op that makes the unified KV cache efficient for multi-sequence inference, and a generic batched tensor copy API.
Saved . A shortened excerpt from GitHub Releases; the original publisher retains authorship.