The pricing floor just dropped again. Tencent launched Hy3, a coding-focused AI model, at $0.14 per million input tokens. That is not a teaser rate or an introductory promotion — it is the listed price, and it lands well below what most frontier models charge for developer-facing workloads.
Here is what the number actually means. At $0.14/M input tokens, a developer running 10,000 iterative coding-agent calls per month pays peanuts compared to GPT or Claude equivalents. That changes the calculus on always-on agent loops — auto-fixing lint errors in CI, generating tests on every commit, or running multi-step refactors that would have been too expensive to leave running overnight.
The catch: input price is only half the equation. Output token cost, context-window limits, and real-world code generation quality matter just as much. A cheap model that hallucinates half its diffs is no bargain. But a credible coding-specialized model from a major lab at this price point puts pressure on every other vendor to justify their per-token rates. Expect the pricing transparency trend to accelerate.
The takeaway: Hy3 resets the low end of coding AI pricing. If quality holds up, it changes how much agentic iteration you can afford to run.