DFlash 2 Released for Qwen 3.8 27B and Muse Glimmer with llama.cpp Support

DFlash 2 has been released for the Qwen 3.8 27B and Muse Glimmer models, accompanied by a pull request to integrate support into the llama.cpp repository. This update brings the second version of DFlash GGUF quants to these specific models. This integration allows users of local LLMs to leverage DFlash's block diffusion speculative decoding for faster inference on consumer hardware. By supporting popular models like Qwen and Muse Glimmer in llama.cpp, it makes high-speed, lossless acceleration more accessible to the open-source AI community. The update is supported by a pull request submitted by GitHub user rerri to the main llama.cpp repository. DFlash works by generating draft tokens in a single forward pass, aligning block-level diffusion predictions with the target model's outputs.