DeepSeek announced on August 6 plans for a significant, across-the-board increase in its API prices1. In a notice, the company warned that the increase could be substantial but did not give the new rates or a date they take effect2.
The new prices are to be set out in a formal notice. DeepSeek's current API prices differ for input and output tokens and for cached and uncached requests. For V4 Flash, one of its two main offerings alongside V4 Pro, output costs 2 yuan per million tokens and input 1 yuan, falling to 0.02 yuan when the input is cached.
Pandaily tied the increase to rising compute costs. V4 Flash recently topped OpenRouter's weekly global ranking of token usage at 7.22 trillion tokens, and on August 4 the V4 Flash API suffered a capacity outage under record inbound traffic.
The move follows an earlier pricing change: a DeepSeek upgrade email at the end of June said the official V4 release, scheduled for mid-July, would bring peak and off-peak pricing, with API prices doubling during weekday peak hours. Other Chinese providers have raised prices too: Zhipu AI CEO Zhang Peng said its API pricing rose 83 percent year over year in the first quarter of 2026.
ANALYSIS Without the new rate card, the size of the increase is unknown; developers who rely on DeepSeek's API face an open-ended cost change until the company publishes its prices.