Same feature, 3 models: $4 Claude vs $0.50 Tencent Hy3 — the cheapest won
tejasghutukade · reddit · 2026-09-15
A developer dogfooding Stageflow ran the same feature request through an identical multi-stage workflow (brainstorm → plan → implement → review → fix → verify → PR) with three models: Claude Sonnet 5 ($4), Qwen 3 Coder ($1.70), and Tencent Hy3 ($0.50) — and ended up merging the $0.50 Hy3 implementation.
His takeaway: the industry may be over-indexing on which model is smartest. With proper task decomposition, intentional context handoff, and output verification, you don't need the most expensive model on every step — models can be treated as specialized workers inside a controlled engineering process. Stageflow is his open-source project for configurable multi-stage agent workflows (tejasghutukade/stageflow).
More from coding & agent
- Andrew Ng: Prompting dies in 6 months, replaced by loops and graphs — Roger_M_Taylor · 2026-09-15
- Stripe Engineer Builds a T-Shirt Factory Factory to Benchmark Dev Tools for Agents — hazelcough · 2026-09-15
- Test: 7 Models Building a T-Shirt Shop Prefer Stale Training-Data APIs Over Latest Docs — hazelcough · 2026-09-15
- Same Test: 0 of 21 Runs Produced a T-Shirt Design I Would Wear — hazelcough · 2026-09-15
- Stripe Engineer: A T-Shirt Shop Is a Useful Minimal Testable Unit of E-Commerce — hazelcough · 2026-09-15
- 7 Models, 21 Runs: Only 7 Produced a Working T-Shirt Shop, Failures Were Silent Bugs — hazelcough · 2026-09-15