Tencent has launched Hy3, a coding-focused AI model priced at $0.14 per million input tokens — a number that should make every team paying premium rates for code generation take a second look. In a market where the big-name models charge multiples of that for similar work, Hy3 is positioning itself as the budget-friendly engine for developers who burn through tokens by the billion.
The model is built for the bread-and-butter of software work: writing functions, refactoring modules, explaining unfamiliar code, and turning natural-language specs into working pull requests. Tencent is betting that for the majority of coding tasks, raw flagship performance is overkill, and a lean model at a fraction of the cost wins on total bill rather than leaderboard bragging rights.
What makes the pricing notable isn’t just the headline rate — it’s the math at scale. A team running thousands of agentic edits a day can see its monthly API spend collapse when the per-token rate drops this low. That lowers the barrier to letting agents handle more of the backlog instead of rationing their calls.
The catch, as always, is quality on the hard problems. Cheap tokens are worthless if the model fumbles architecture decisions or subtle bugs. But if Hy3 holds up on routine engineering, it sets a new price floor the rest of the field will feel. For cost-conscious shops, the calculus just shifted.
Before you buy any coding agent subscription — compare prices and find exclusive deals at aiFiesta.