Chantless specialist extraction: 8B model beats big model at function calling by 107.8%
tokenbender · x · 2026-10-14
tokenbender's Chantless article details a specialist-extraction methodology: define what doing a task well means, then shape a model around that standard instead of shipping all weights. Benchmarks show Qwen3-8B hitting 66.5% on BFCL single-call (100.4% of the big model, and 107.8% in a smaller config), Qwen3.8-27B scoring 77.8% on Terminal-Bench 2.1 pass@1 (104.1%), 96.67% on AIME 2025 avg@8 (97.5%), and 0.725 on RedactionBench (102.5%) — suggesting task-tuned small models can match or beat full-size models.
Related event: Chantless: Taste-Grown Custom Models with 8B Beating Giants by 107.8%(3 posts)→
More from coding & agent
- When building is 10x faster, knowing when to slow down matters 100x more — brandon_galang · 2026-10-14
- Matt Pocock: compacting the orchestrator session is safe because its task is simple — mattpocockuk · 2026-10-14
- Building games with AI? Start with a graybox playable before final art — chongdashu · 2026-10-14
- Gemini 4 Argon goes live on Vercel AI Gateway at 50% off as rauchg hails coding agents as AI's killer app — AccBalanced · 2026-10-14
- Dev shares slash-command combo: prototype with /ce-prototype, then build with /lfg — kieranklaassen · 2026-10-14
- Kevin Rose hails Claude 'Projects' beta as a 'beautiful orchestration of sub-agents' — joellewenstein · 2026-10-14