llama.cpp b9019 Release Enhances Model Flexibility

How it developed

1. May 5, 2026llama.cpp ReleasesWhere it started

   llama.cpp b9019 Release Enhances Model Flexibility

2. May 10, 2026llama.cpp Releases

   BF16 Support Added to llama.cpp SYCL Backend

   llama.cpp SYCL backend adds BF16 support to GET_ROWS, eliminating GPU-to-CPU transfers for models like Gemma4.