DeepSeek raises API prices up to 4x

DeepSeek will raise API prices by more than 4x at peak hours starting August 16, according to pricing documentation the company posted on its own site and reporting confirmed by Fortune and Bloomberg. V4-Flash peak-hour output pricing rises from $0.28 to $1.32 per million tokens; V4-Pro rises from $0.87 to $3.96. Both models keep a 50% off-peak discount, so V4-Flash off-peak lands at $0.66 and V4-Pro at $1.98, still a steep jump from today’s flat rates.

DeepSeek says the dynamic peak/off-peak split is meant to “allocate resources more reasonably” and push developers to shift non-urgent workloads into quieter windows, not simply to raise margin. Even after the increase, DeepSeek remains far cheaper than frontier labs, Bloomberg notes Anthropic’s Claude pricing runs near $50 per million tokens by comparison, so the company’s core pitch as the low-cost option is intact.

The timing lines up with DeepSeek’s own IPO preparations, reported to be underway for as soon as this year, and comes as OpenAI and Anthropic have both been cutting prices to compete for developer volume.

What it means for you

If cost is why you route traffic to DeepSeek, check whether your workload actually needs peak-hour speed, or whether batching, queuing, or scheduling calls into off-peak hours saves you most of the increase for free. DeepSeek’s V4-Pro just reached general availability, so this is also a sign the company is done treating price as a one-way lever now that it has a flagship worth charging more for. Pair it with GLM-5.3’s launch the same week: the cheap, open tier of this market is still growing fast, but it’s no longer free of the same tradeoffs everyone else makes.