A fine-tuned 4B Qwen in 2.6 GB beats GPT-5.6 on a transit-kiosk agent benchmark, and the PEFT gain vanishes by 27B.

MetroLLM-Bench is 955 cases across six real metro systems of 37 to 414 stations, requiring structured tool calls and a machine-renderable terminal state. On the 238-case held-out split the 4B student scores 91.3 on Tier 1 against GPT-5.6's 90.6 and 90.0, at Q4_K_M. The gain over base shrinks monotonically from +7.03 points at 2B to -0.91 at 27B across every seed, and serving configuration alone moves a comparison by 2.7 points. For a bounded tool-calling domain, small and fine-tuned wins on cost and latency and there's no reason to go past 4B.