Chinese AI DeepSeek cuts off-peak pricing for developers by up to 75%

Chinese AI startup DeepSeek introduced discounted off-peak pricing on Wednesday for developers looking to use its AI models to build their own products, its website showed, a move that could put pressure on rivals in China and overseas to cut prices.

As investors worried that its arrival could threaten current AI market leaders, DeepSeek's low-cost AI models triggered a major sell-off in global equity markets last month.

The Hangzhou-based company said on Wednesday that between 1630GMT and 0030GMT, the cost of using its API, a platform that allows developers of other apps and web products to integrate its AI models, would be up to 75% cheaper.

Usage costs during this time frame for the API of the R1 and V3 models would be 75% and 50% cheaper, respectively, according to a table on DeepSeek's website.

While the company calls this time frame "off-peak" as it starts at 0030 and ends at 0830 in Beijing, it encompasses daytime hours in Europe and the United States, where DeepSeek's cheap but powerful models triggered a sell-off in tech stocks.

Wednesday's price discounts are the latest in a series of moves by DeepSeek that have shaken the AI industry, in China and abroad. The company is now accelerating the launch of a successor to January's R1 model, people familiar with the company have said.

The company's open-source ethos and low pricing first triggered a domestic AI model price war last May, when it released V2, the predecessor to the model powering its globally popular AI assistant.

Since the release of its AI assistant last month, ChatGPT makers OpenAI cut prices, while Google's Gemini has introduced discounted tiers of access.