LaunchAI ModelsNovember 17, 2026

Hugging Face releases 300MB LoRA adapter for Qwen3.8-27B on OpenCode

Read original source →huggingface.co

The RL-trained LoRA (r=32, ~300MB) runs on a frozen Qwen3.8-27B inside the OpenCode coding agent and is the second harness adapter after the Pi release. Reported gains are measured on SWE-bench Pro's hard 51-task subset; a GGUF build is included for local inference.

How this story unfolded

1 day · 1 report · 4 community posts · from Nov 18

  1. Nov 18
  2. Nov 19

More stories today

Open the live feed