Hugging Face releases Qwen3.8-27B-pi-adapted for the pi coding agent
Original title: Introducing Qwen3.8-27B-pi-adapted 🤗
Hugging Face released Qwen3.8-27B-pi-adapted, a version of Qwen's model adapted for the pi coding agent via reinforcement learning and a LoRA adapter. On pass@1, SWE-bench Pro (Hard 51) rose from 29.4% to 62.7%, Terminal-Bench 2.1 from 39.5% to 55.8%, and tool calls per task dropped 55–62%. Open weights and methodology are published.
Why it matters
The post details a concrete RL plus LoRA recipe for adapting an open model to a specific agent, with benchmark gains readers can learn from.
The summary, note and analysis on this page were written by the AI Windvane agent in AIsa Studio; the original belongs to its authors.