DeepSeek price increase will hit API services, with rates still unknown
DeepSeek says it will raise API prices soon, but has not disclosed the new rates, start date or which models will be affected.
By Bianca Rossi · Entertainment Editor
3 min read
A DeepSeek price increase is coming for the company’s API services, and developers are still waiting for the numbers. The Chinese AI company told users its overall API pricing would rise in the near future and called the expected increase significant, according to reports by Bloomberg and the South China Morning Post.
The notice did not give an effective date, a percentage increase or a new price list. It also did not spell out whether every API model will be covered. For customers budgeting for software that relies on DeepSeek, the practical message is clear: costs are set to rise, but the final bill cannot yet be calculated.
What will DeepSeek’s API prices be after the increase?
DeepSeek has not said. Bloomberg reported that the company urged users to plan their usage, while SCMP said specific plans would be announced later. The available reporting identifies API services as the scope of the notice, so it does not establish changes to DeepSeek’s consumer app, subscriptions, self-hosted models or other products.
An API, or application programming interface, lets another app or service send requests to an AI model. Developers are generally billed by tokens: input tokens cover the material a model processes, while output tokens cover the material it generates.
What DeepSeek charges now
Bloomberg listed DeepSeek V4 Flash’s current price at $0.14 per million input tokens and $0.28 per million output tokens. Those are the rates that will be the reference point for developers once the company publishes its revised pricing.
- Known: DeepSeek has signaled a significant increase to overall API pricing in the near future.
- Known: V4 Flash is currently priced at $0.14 per million input tokens and $0.28 per million output tokens, Bloomberg reported.
- Unknown: the new dollar rates, the size of the increase and the date it takes effect.
- Unknown: the model-by-model reach of the changes and whether non-API products will be affected.
The current rates have stood well below figures Bloomberg cited for other AI services. Moonshot’s Kimi K3 was listed at $3 per million input tokens and $15 per million output tokens, while Anthropic’s Fable 5 was listed at $10 and $50. Those figures are pricing context, not proof that the services are directly comparable or that DeepSeek will remain cheaper after its increase.
The announcement followed the release of DeepSeek-V4-Flash-0731, a lightweight V4-series model with 284 billion parameters, SCMP reported. The outlet said demand for low-cost AI models had been rising, but DeepSeek did not give a reason for the planned hike in the notice.
Bloomberg has separately reported that DeepSeek intends to pursue 1 gigawatt of AI computing power, including a planned data center in Inner Mongolia and potentially leased capacity. That expansion may require substantial funding, Bloomberg said, though the company has not linked it publicly to the pricing decision.
This story draws on original reporting from Mashable.