llama.cpp version b9723 introduces support for Qwen3.5 and Qwen3.6 models via Eagle3. The release includes deferred boundary checkpoints restore for hybrid models and updates to API and naming conventions. Binary builds are available for macOS, Linux, Android, Windows, and openEuler platforms, with options for CPU, Vulkan, OpenVINO, SYCL, and ROCm.
llama.cpp release b9723 adds support for Qwen3.5 and Qwen3.6 Eagle3
from English