Solo fine-tune of Qwen3.8-27B-pi fixes effort ordering, saves 41% tokens
victormustar · x · 2026-10-01
A solo developer fine-tuned Qwen3.8-27B into pi on rented GPUs for the Pi coding agent, fixing the base model's broken effort ordering (low often reasoning longer than medium). Two-stage recipe: SFT on curated successful Pi sessions, then GRPO RL with a success-conditioned reward enforcing effort ordering. Result: non-decreasing reasoning tokens and pass rates from low→xhigh on Terminal-Bench 2.1, GPQA Diamond and SciCode; pi at medium matches base at xhigh (67/89) with 41% fewer output tokens, and 23% fewer on SciCode xhigh. BF16, FP8 (30.4 GB) and 17 GGUF variants released.
Related event: Dev Fine-tunes Qwen3.8-27B-pi to Fix Reasoning Effort Ordering(2 posts)→
More from coding & agent
- Generic API-catalog MCP server vs one tool per job: a context-bloat experiment — MandeepKaur9407 · 2026-11-16
- Berkeley paper: LLMs answer with stale values even when the update is in context — ombre-ai · 2026-11-16
- skills-gate: one gate skill fronts 148 agent skills at just 389 tokens — Distinct-Pie2389 · 2026-11-16
- Open-weight Karpo-Decision-31B tops Decision Index at 73.9, one forward pass per answer — CodeByPoonam · 2026-11-16
- shadcn boosts @uisdk: one-shot UI rebuild that's instantly E2E-testable by AI — shadcn · 2026-11-16
- AI agent deletes 43,599 lines of OpenClaw test code in one PR — vincent_koc · 2026-11-16