Skip to content
AI 24 UpdatesAI 24UpdatesWhat changed. Why it matters to you. What to do.

Hugging Face releases RL-trained harness adapters for Qwen3.8-27B in pi, Hermes Agent and OpenCode

Version 1

Share on X

In Brief

  1. Hugging Face — harness adapters: RL-trained LoRAs for Qwen3.8-27B in pi, Hermes Agent and OpenCode
Hugging Face blog image for Harness Adapters: LoRA adapters for Hermes, pi and opencode trained with RL on Qwen3.8-27B
Image: Hugging Face

Details

Hugging Face — harness adapters: RL-trained LoRAs for Qwen3.8-27B in pi, Hermes Agent and OpenCode

Hugging Face

Hugging Face trained a LoRA adapter for Qwen3.8-27B inside each of three agent harnesses and merged them into one super-adapter, all released on the Hub.

  • Hugging Face released one LoRA adapter for Qwen3.8-27B per agent harness — pi, Hermes Agent and OpenCode — each trained with reinforcement learning inside the harness itself.
  • Datasets, training and agent rollouts ran on Hugging Face Jobs with TRL's AsyncGRPOTrainer, OpenEnv and Harbor; training totalled 598 steps and 167M tokens.
  • Each adapter beats the base model under its own harness on Terminal-Bench 2.1 and the 51 hardest SWE-bench Pro tasks; under pi on SWE-bench Pro, the score rises from 40.0% to 64.0%.
  • The three LoRAs are also merged into a single rank-96 super-adapter (each update at weight ½) that runs unchanged under pi, Hermes Agent and OpenCode.
  • The adapters, GGUF files, RL tasks and the safeagents plugin are published in a Hugging Face collection.
Sources:Hugging Face blogHugging Face collection

Subscribe for updates

Subscribe to get AI 24 Updates by email. You'll receive a confirmation email first.

We use your email address only to send the AI 24 Updates newsletter. You will get a confirmation email first, and every newsletter has an unsubscribe link. Subscriptions are managed by Buttondown (USA). See our Privacy Policy.