Qwen3.8 27B is a 27B dense instruct model served on vLLM with FP8 quantization, tuned for concise general-purpose chat.
