LoRA adapter boosts Qwen3.8 coding pass rates to 65% on SWE‑bench Pro
Hugging Face’s Qwen3.8‑27B‑pi‑adapted LoRA raises coding success from 39% to nearly 65% on key benchmarks.
Hugging Face released a new LoRA adapter named Qwen3.8‑27B‑pi‑adapted on 18 November 2026.
The adapter is Apache‑2.0 licensed and was trained with reinforcement learning for the pi coding agent.
On the SWE‑bench Pro hard‑51 benchmark, pass@1 improved from 39.2% to 64.7%.
On Terminal‑Bench 2.1, the pass rate rose from 39.5% to 55.8%.
These gains show the adapter’s ability to enhance Qwen3.8’s code‑generation performance without changing the base model.
What it means for you
Developers can obtain higher coding accuracy by applying the free LoRA adapter to an existing Qwen3.8 deployment, avoiding the need for larger paid models.
Sources
- New LoRA Adapter Lifts Qwen3.8 Coding Success From 39% to 65%, MarkTechPost, 18 Nov 2026
- harness-hub/Qwen3.8-27B-pi-adapted · Hugging Face, huggingface.co (official), 18 Nov 2026
Written automatically from the sources above, then checked by our code: every number and name must appear in them, and no sentence may copy a source. Spotted a mistake? Report it.
Which AI can you use for free today?
Every free AI model, ranked by blind human votes, with the memory and speed it needs on your own GPU.