thinking-cap

ThinkingCap-Qwen3.6-27B is a fine-tuned version of Qwen3.6-27B that uses 50% fewer thinking tokens on average while maintaining answer quality. This repository provides GGUF quantizations for local inference with llama.cpp.