DeepSeek announces steep price increase
DeepSeek, the Hangzhou-based artificial intelligence lab, said on Thursday that it will raise the cost of its flagship V4 models from 1 August. During peak hours the new rate for the DeepSeek-V4-Flash model will be $1.32 per million output tokens, up from $0.28. The off-peak price will be half of that. The V4-Pro model will move from $0.87 to $3.96 per million tokens at peak times, with a similar off-peak discount.
Why the new pricing matters
The adjustment brings DeepSeek's fees closer to those of larger rivals such as Anthropic, whose Fable 5 service costs $50 per million tokens. While still cheaper, the hike narrows the gap that had made DeepSeek a low-cost alternative for developers. The company says the dynamic pricing is intended to allocate resources more reasonably and to encourage users to shift workloads to less congested periods.
Implications for the AI market
The move adds fuel to an ongoing debate about the affordability of high-performance AI tools. Critics have warned that rising costs could create a "death zone" where only the most expensive or capable models survive. The price change also arrives as DeepSeek prepares for a major fundraising round and a possible initial public offering later this year.
What comes next
Liang Wenfeng, DeepSeek's founder, will now need to balance investor expectations with the capital demands of expanding computing infrastructure. The new rates are set to take effect on 16 August, giving developers and enterprises a short window to adjust their budgets or consider alternative providers such as OpenAI or Anthropic.