[Amid Price Hike Wave, DeepSeek Drives AI 'Price War': 1 Million Tokens for 0.25 RMB] DeepSeek has recently launched a special 75% discount for its latest DeepSeek-V4-Pro model API. The input (cache hit) price is 0.25 RMB per million tokens, the input (cache miss) price is 3 RMB per million tokens, and the output price is 6 RMB per million tokens. The promotional period ends on May 5. According to Open Router data, the newly released GPT-5.5 Pro, a renowned overseas AI large model, has a weighted average input price of $30 per million tokens and an output price of $180 per million tokens, making its input price over 700 times higher than that of DeepSeek V4 Pro. Additionally, the GPT-5.5 Standard version has a weighted average input price of $5 per million tokens and an output price of $30 per million tokens. Other models, such as the Anthropic Claude Opus series, OpenAI GPT-5.4, and Google's Gemini 3.1 Pro series, have output prices ranging between $12 and $25, all showing a significant gap compared to the adjusted prices of DeepSeek V4 Pro. Notably, against the backdrop of a collective price increase in computing power across the AI industry this year, DeepSeek V4 has once again upheld its philosophy of 'AI price reduction,' further intensifying the industry's 'price war.' (The Paper)