DeepSeek, the Chinese AI lab that became globally famous for proving that strong models could be served at dramatically low prices, is now doing the opposite of what made it famous: it is raising prices. On August 13, the company announced that API rates for its new V4 Pro flagship and V4 Flash models will rise by 50% to as much as 1,100% on some workloads, taking effect August 16.
Under the new pricing, V4 Pro output tokens will cost $3.96 per million at peak and $1.98 per million off-peak, versus $0.87 per million today; cache-hit input pricing rises even more sharply, which is where the largest increases — around 1,100% — occur. The company is also introducing peak and off-peak billing, with off-peak rates 50% lower, a first for DeepSeek's API.
The move is a striking pivot for a company whose rise was built on undercutting Western rivals. It signals that Chinese labs are moving from pure price disruption toward monetizing premium performance: reliability, coding ability, reasoning and throughput now command a premium. It also arrives as OpenAI and Anthropic cut prices on their own models in response to intensifying competition, creating the unusual situation of premium Chinese models getting more expensive while some American alternatives become cheaper.
Industry analysts note that the shift reveals an emerging reality: inference prices may not move in one direction forever. Better models consume expensive compute, and providers with strong demand will test how much customers are willing to pay. Even after the increase, DeepSeek remains inexpensive relative to several frontier competitors — but the era of AI as a guaranteed deflationary commodity may be ending, one API at a time.
