Amid the price hike wave, DeepSeek drives an AI "price war": 0.25 yuan for 1 million tokens input

DeepSeek’s latest release, the DeepSeek-V4-Pro model API, is now available at a 2.5-fold price discount: when the input ( cache hits, the price is 0.25 yuan per million Tokens; when the input ) cache misses, the price is 3 yuan per million Tokens. The output price is 6 yuan per million Tokens, and the promotional period ends on May 5. Meanwhile, according to Open Router data, the newly released well-known overseas AI large model GPT-5.5 Pro has a weighted average input price of $30 per million Token and a weighted average output price of $180 per million Token—already more than 700 times higher than DeepSeek V4 Pro in terms of input price. In addition, the weighted average input price of the GPT-5.5 standard version is $5 per million Token and the output price is $30 per million Token. The output prices of large models including the Anthropic Claude Opus series, OpenAI GPT-5.4, and Google Gemini 3.1 Pro series range from $12 to $25, leaving a much larger gap than the re-priced DeepSeek V4 Pro. Notably, against the backdrop of an across-the-board increase in AI industry computing power prices this year, DeepSeek V4 once again fully implements the “AI price reduction” philosophy, further escalating the industry’s “price war.” (The Paper)