Qwen3.8-27B-Humanlike-Chat 2.0: texts like a human, now with tool calls and better instruction following

Qwen3.8-27B-Humanlike-Chat 2.0, a LoRA model, has been released, building on the success of its predecessor which garnered over 700 upvotes and 44k downloads. This new version significantly improves human-like text generation, tool calls, and instruction following. In tests, Model Judge found it indistinguishable from a real person 50% of the time, and it was preferred over the base model in 16 out of 16 multi-turn chats. Safetensors for vLLM and SGLang are available in GPTQ-Int4, FP8, and BF16 formats.