Chinese startup DeepSeek has decided to make a 75% discount on its latest flagship AI model, the DeepSeek V4 Pro, permanent. The promotional offer, which was due to expire on May 31, 2026, is now fixed as the base rate.
Chinese startup DeepSeek has decided to make a 75% discount on its latest flagship AI model, the DeepSeek V4 Pro, permanent. The promotional offer, which was due to expire on May 31, 2026, is now fixed as the base rate.
This solution significantly reduces the cost of using the technology for corporate clients and developers working with large amounts of data. As Engadget reports with reference to official data on the company’s website, the new tariffs make DeepSeek one of the most affordable large language models on the global market.
According to the updated pricing, prices for using the DeepSeek V4 Pro API now range from $0.003625 to $0.87 per million tokens (depending on the type of request — cached input or output). Previously, this range was from $0.0145 to $3.48. Thus, the cost of working with the model has fallen exactly four times, which will allow businesses and developers to significantly reduce the costs of generating millions of tokens daily.
This decision was made just a month after the release of the V4 line of models (Pro and Flash). At the time, the company stated that it was striving to open «the era of cost-effective context lengths of 1 million tokens.» Now, the deep discount creates serious competition for American industry leaders — in particular, OpenAI’s GPT-5 models and Google’s recently presented Gemini 3.5 Flash.
