Key Info

Hugging Face has released Qwen3.8-27B-pi-adapted, a model adapted specifically for the pi coding agent using reinforcement learning and a LoRA adapter, with open weights and methodology available.

Highlights

  • SWE-bench Pro (Hard 51) pass@1 improved from 29.4% to 62.7%
  • Terminal-Bench 2.1 pass@1 improved from 39.5% to 55.8%
  • 55–62% fewer tool calls per task
  • Open weights and methodology are provided