qi's EXACT verbatim prompt · "build the full business in one shot" · 4 local models · ranked by valid+complete+coverage
✗ deepseek-r1:8b · 5.2GB · 9m 39s
TRUNCATED — did not ship
Ran 9.5 min, blew its token budget on <think> overhead, HTML never closed. Not deployable.
✗ qwen3-coder:30b · 18.6GB · 10m 2s
FAILED — did not ship
Ran 10 min, produced 15 bytes of garbage on the long prompt. Choked. Not deployable.