HF's harness adapters lift Qwen3.8-27B from 39% to 55% on Terminal-Bench 2.1 with half the tokens
SergioPaniego · x · 2026-11-22
Hugging Face released harness adapters, a technique to adapt a model to your coding harness to improve both performance and cost.
- With an adapter, Qwen3.8-27B in Pi goes from 39% to 55% on Terminal-Bench 2.1 while using half the tokens.
- Led by Younes Belkada with the HF open-source team; technical article and code are published.
A drop-in way to gain benchmark score and cut inference cost without swapping the underlying model.
More from coding & agent
- LangChain founder on 'knowledge factories': better enterprise knowledge, not better models — hwchase17 · 2026-11-22
- Pi Bot open-sources multi-agent teamwork as a group chat, with sandboxed local execution — lxfater · 2026-11-22
- Ransomware gangs wired Claude Opus and Kimi into attacks: 32 MCP servers, $4.5K AI-driven ops — cyb3rops · 2026-11-22
- ArcKit v6.25.0 brings 190 architecture governance Agent Skills to Google Antigravity — mcraddock · 2026-11-22
- A 6-step recipe to build your own knowledge factory with Obsidian plus a cloud agent — hwchase17 · 2026-11-22
- Claude played three roles in the Riemann Hypothesis proof: ideation, verification, explanation — Prestigious_Bag_5349 · 2026-11-22