DeepSeek warns of a significant API price increase, reversing its low-cost strategy

DeepSeek warned that it plans to raise its application programming interface (API) prices “in the near future, with a significant increase expected,” according to a notice on its API pricing page. The company gave no effective date or specific new rates, leaving customers to price in an increase they cannot yet size.

The warning is a notable reversal for a company that built its reputation on undercutting rivals. DeepSeek-V4-Flash currently costs $0.14 per million input tokens and $0.28 per million output tokens, among the cheapest rates for a capable model.

The timing points to capacity, not just strategy. The notice follows the July 31 release of DeepSeek-V4-Flash-0731, a 284-billion-parameter lightweight model, and comes amid a reported surge in demand for the company’s low-cost models and tight compute capacity inside China, where access to advanced AI chips is constrained by export controls.

The move complicates DeepSeek’s low-price story. Only weeks earlier the company had announced a price cut tied to context caching on disk; a broad increase now would push in the opposite direction.

For developers who standardized on DeepSeek to keep inference bills down, the open question is how far prices rise and when. The company has not said.