DeepSeek introduces double peak-hour pricing, a first in China's AI price war

In April 2026, DeepSeek, the Chinese AI startup that triggered the price collapse in the market for language model APIs, notified developers that it is introducing dynamic pricing: during peak load hours, token costs for its V4 series models will double.

What does the new pricing mean for developers?

According to the South China Morning Post, which reviewed the company's official notice, the peak-hour surcharge applies to the V4 line, DeepSeek's latest model family. The company does not disclose the exact range of "peak hours" or the algorithm used to calculate the price.

Key facts about the new pricing regime:

• Peak-hour surcharge: ×2 on the base price per token
• Applies to DeepSeek V4 series models
• The change affects only API customers, not end users of apps
• No competing Chinese AI provider has made a similar move so far
• The details became known through the South China Morning Post

Surge pricing has long been standard practice in cloud services (Amazon AWS, Google Cloud) and ride-hailing apps (Uber, Yandex Taxi). It has been almost unused in the market for language-model APIs: major providers, from OpenAI to Alibaba, have historically kept a fixed price per token regardless of infrastructure load.

Why is this move surprising?

The irony is that DeepSeek was the one that fired the first shot in the price war. In early 2025, the company released open, cheap models that, at competitive quality, cost an order of magnitude less than GPT-4-class solutions. Alibaba Cloud, Baidu, ByteDance and a number of other Chinese AI labs were forced to slash prices in response, and some temporarily offered free access to their APIs.

South China Morning Post commentators call the situation paradoxical: the company that launched the race to the bottom on price became the first to step the other way.

The peak-hour surcharge is an indirect admission of reality: subsidized prices have a limit. When computing capacity runs at its limit, a flat price list stops covering variable costs. Double pricing solves two problems at once: it smooths peak demand by encouraging customers to shift tasks to less busy times, and it improves service quality for customers willing to pay for priority access.

What this means

DeepSeek is the first participant in China's AI price war to adopt variable pricing. It sets a precedent that others may follow: cheap tokens cannot be subsidized indefinitely, and dynamic pricing in AI APIs is only a matter of time.

For development teams building products on DeepSeek V4, surge pricing is a new variable in budgeting. The sensible approach is to move tasks without strict latency requirements to off-peak hours, and to factor the double rate into calculations for realtime requests.

Frequently asked questions

Which DeepSeek models are affected by the increase?

Surge pricing applies to the DeepSeek V4 series, the latest family at the time of the announcement. Models from earlier generations are not mentioned in the company's official notice.

Has any competitor done something similar?

According to the South China Morning Post, no Chinese AI provider had introduced dynamic pricing before DeepSeek. In the Western market, OpenAI, Anthropic and Google also stick to a fixed price per token.