DeepSeek raises prices for V4-Flash AI model, shaking market

Chinese AI lab DeepSeek announced an upcoming increase in the fees for its V4-Flash model and other API services. The company, which currently charges $0.14 per million input tokens and $0.28 per million output tokens, said the new rates will be higher, with peak‑hour pricing already set to double during two daily windows in Beijing time.

The price hike marks a shift for a service that has been marketed as a low‑cost alternative to offerings from OpenAI, Anthropic, Google and others. DeepSeek’s V4‑Flash, built on a mixture‑of‑experts architecture with 284 billion parameters and a one‑million‑token context window, has become the most used AI model worldwide, recording 7.22 trillion tokens processed in a week and 8 trillion tokens on a single day, according to OpenRouter data. Half of that usage was free, the rest paid.

DeepSeek also disclosed plans for a 1‑GW AI data centre in Inner Mongolia and potential financing and IPO activities, underscoring its rapid growth. The announced price changes could affect thousands of developers and enterprises that have relied on DeepSeek’s cheap pricing to run large‑scale automations and AI‑driven products.