DeepSeek built its name by being the cheap option. That chapter is closing.
The company is raising prices across its V4 lineup, with the increases taking effect on August 16. Peak hour pricing on DeepSeek V4 Flash is jumping to $1.32 per million output tokens, more than four times what it cost before. Off peak hours get a discount at half that rate, DeepSeek’s attempt at spreading out demand instead of just charging everyone more across the board.
The company says it is moving to allocate resources more reasonably, which is a polite way of saying demand has outrun the hardware. DeepSeek’s models have been getting real traction since the V4 launch, and running frontier scale inference at rock bottom prices was never going to hold once enough people showed up to use it.
There is a bigger story sitting underneath this one. The assumption that Chinese open models would keep undercutting Western pricing forever was always a little too neat. Compute is compute, and somebody pays for the GPUs whether the checkpoint is open or closed. DeepSeek finding religion on pricing is a sign the whole industry is running into the same wall, just from different directions.
None of this means DeepSeek stops being competitive. Off peak pricing is still aggressive, and the model itself keeps improving. It just means the free lunch phase of this whole AI race is wrapping up everywhere, not only in San Francisco.
For anyone building on top of these models, this is the actual lesson. Do not architect your whole product around whichever API happens to be cheapest this month. Prices move as demand catches up to supply, and the labs that looked like charities a year ago are all discovering the same math everyone else already knew. That is not a reason to panic, it is just a reason to build with a little more room to move.
——
Follow: @Ali Demi
Book your free AI clarity call, NOW!
https://buff.ly/TpWy277
——
Sources:
https://x.com/deepseek_ai/status/2087864589895798968
https://finance.yahoo.com/technology/ai/articles/deepseek-increases-prices-ai-services-125256361.html
https://www.engadget.com/2236912/deepseek-ai-models-get-four-times-pricier/
Repost this. Thanks.

