DeepSeek increases prices for AI services by multiple times

DeepSeek is raising the price of its flagship V4 models, pushing the Chinese lab much closer to the kind of pricing seen at bigger AI rivals. The new rates kick in on Aug. 16 and, in peak hours, will jump by more than four times from current levels, according to a post on the company’s website Thursday.

For the DeepSeek-V4-Flash model, users will pay $1.32 for 1 million output tokens during peak hours and half that when demand is lower. That compares with the current $0.28 per million tokens. The V4-Pro model is also getting a steep reset: $3.96 per million tokens at peak hours, and half that off-peak, up from $0.87.

DeepSeek says it is revising pricing “to allocate resources more reasonably,” and the pattern is meant to steer developers and enterprises toward less crowded times. The company had warned last week that higher prices were coming, but it hadn’t said by how much.

Even after the increase, DeepSeek is still cheaper than some major competitors. Anthropic PBC’s Fable 5 service is priced at $50 for the same amount of output. But DeepSeek’s move has already fed a bigger argument about how much high-performing AI should cost, and whether the market has room for models that undercut the incumbents so aggressively.

That debate is now spilling into the business plans of OpenAI and Anthropic, with questions swirling around their IPO prospects and profit expectations. DeepSeek, meanwhile, is in the middle of a large fundraising process and has started preparing for a possible IPO as soon as this year, according to Bloomberg. Founder Liang Wenfeng is about to find out what happens when price disruption meets the bill for serious computing power.