Chinese company DeepSeek will significantly increase the cost of using its flagship V4 model and abandon its previous ultra-low tariffs. The new prices will take effect on August 16 and will depend on the load on the computing resources.
Photo credit: John Cameron/Unsplash
It is understood Bloombergthe cost of 1 million DeepSeek-V4-Flash output tokens during peak hours will be $1.32, compared with $0.28 before. Tickets are half price the rest of the time. The more advanced model, DeepSeek-V4-Pro, will also become more expensive: the price will rise from $0.87 to $3.96 per 1 million tokens during peak periods, and will be half as low outside of this period.
The company explained the tariff changes as a need to allocate computing resources more rationally, and the dynamic pricing model should encourage developers to shift resource-intensive tasks to less busy periods. However, even after the price increase, DeepSeek’s rates will still be lower than some of its larger competitors. For example, purchasing a similar amount of Claude 3.5 Sonnet tokens from Anthropic costs $50.
Bloomberg noted that the sharp change in tariffs was due to growing concerns about the price of high-performance artificial intelligence models. Previously, the unusually low cost of DeepSeek’s service sparked discussions about the so-called “DeepSeek death zone,” where more expensive or less powerful models could lose competitiveness.
At the same time, according to previous reports by 3DNews, DeepSeek is preparing for large-scale financing and considering an IPO this year, which is why its founder Liang Wenfeng must not only consider the pace of business expansion, but also the growing cost of computing infrastructure.
If you find an error, select it with your mouse and press CTRL+ENTER.