A community developer has released multiple uncensored GGUF-format models, including LongCat-Flash-Lite-Sparse with MTPs and LSAs, Qwen3.8-27B and Qwen3.5-122B-A10B both with MTPs, and Qwen3-Coder-Next and Laguna-S2.1 with vision capabilities. The LongCat-Flash-Lite-Sparse release required the most effort, involving the creation of Heretic support from scratch and custom llama.cpp modifications. The developer also shared links to their llama.cpp fork with LongCat-Flash-Lite support and the J-Wash enhanced fork.