LoRA adapters tuned per harness double Qwen3.8-27B agent benchmark scores
yb2698 · x · 2026-11-17
Instead of arguing about which (model, harness) pair gives the best vibes, Lewis Tunstall's team trained harness adapters: one LoRA per agent harness, swappable at runtime.
The first adapter targets Qwen3.8-27B on the Pi harness, with big gains:
- SWE-Bench Pro (Hard 51 subset) pass@1: 29% → 63%
- Terminal-Bench 2.1 pass@1: 40% → 56%
- 55-62% fewer tool calls per task
They say the era of harness-specific adapters has begun, with more adapters coming in the next days.
Related event: Harness Adapters launch: LoRA doubles Qwen3.8-27B coding scores(4 posts)→
More from coding & agent
- Cloudflare launches x402 Monetization Gateway for AI agents; Solana handles 70% of x402 txns — kleffew94 · 2026-11-17
- Mnemos x NousResearch launches agent memory engine claiming 2x gains over Claude Code — RileyRalmuto · 2026-11-17
- How Revid doubled its ChatGPT-sent signups from 355 to 781 per week in 10 weeks — tibo_maker · 2026-11-17
- FormMCP: one MCP connection to a family of desktop apps, tools loaded on demand — nodemode_ai · 2026-11-17
- Harness Adapters: RL-trained LoRA lifts Qwen3.8-27B on Pi from 29% to 63% on SWE-bench Pro — LysandreJik · 2026-11-17
- Pi harness adapter ships as a LoRA runnable on vLLM, llama.cpp, and MLX — yb2698 · 2026-11-17