DeepSeek V4 model pricing significantly reduced, 97% cheaper than OpenAI GPT-5.5

Foreign Media: DeepSeek's New Model V4 Prices 97% Lower Than OpenAI's GPT-5.5

DeepSeek has significantly reduced the pricing of its AI models. The API "input cache hit" cost for the latest flagship V4 series has been slashed to one-tenth of the original price, with the lowest input cost now approximately $0.14 per million tokens, and this rate is permanently valid.

To promote the recently launched V4-Pro, the company is offering an additional 25% discount until May 5th, bringing the price down to just $0.0036 per million input tokens. In comparison, OpenAI’s GPT-5.5 cache input pricing stands at $0.50 per million tokens. Since input in conversations is typically three times the output, when calculated comprehensively, each conversation with GPT-5.5 costs 32 times more than DeepSeek V4—meaning DeepSeek is about 97% cheaper than OpenAI.

This aggressive pricing strategy stands in stark contrast to recent price hikes by domestic large models such as Kimi K2.6 and Zhipu GLM-5.1. DeepSeek’s move to cut prices against the trend may trigger a new round of price wars in China’s foundational model market, attracting more enterprise clients, developers, and AI agent users.

Original Source: toutiao.com/article/1863725639586819/

Disclaimer: This article represents the personal views of the author.