DeepSeek prices V4 at 97% below GPT‑5.5

DeepSeek cut prices on its new V4-Pro model on April 27, offering developers a 75% discount and reducing cached-prompt fees across its API lineup. (reuters.com) The Hangzhou-based company said the V4 family includes a higher-end Pro model and a lighter Flash model released the previous week. It said the promotional V4-Pro discount runs through May 31, 2026, after extending an earlier May 5 deadline. (api-docs.deepseek.com) (reuters.com) DeepSeek also cut input cache-hit prices for all models to one-tenth of their launch levels, with the change taking effect at 12:15 UTC on April 26. On the official pricing page, V4-Pro cache-hit input is listed at $0.003625 per million tokens during the promotion. (api-docs.deepseek.com) A token is a small chunk of text that application programming interfaces bill by the million. A cache hit means the model can reuse repeated prompt text instead of processing it from scratch, which lowers costs for chatbots and agents that send the same instructions over and over. (openai.com) (api-docs.deepseek.com) That pricing puts DeepSeek far below OpenAI’s current flagship list rates. OpenAI’s pricing page lists GPT-5.5 cached input at $0.50 per million tokens, standard input at $5.00, and output at $30.00. (openai.com) South China Morning Post reported DeepSeek’s discounted V4-Pro cache-hit price is about 97% below GPT-5.5’s cached-input rate. Reuters separately reported the company is using discounts to push adoption as competition intensifies in China’s artificial intelligence market. (scmp.com) (reuters.com) DeepSeek launched V4 days after OpenAI released GPT-5.5 on April 23. OpenAI describes GPT-5.5 as a frontier model for coding and professional work, while DeepSeek is competing on both price and compatibility with OpenAI-style application programming interfaces. (openai.com) (api-docs.deepseek.com) Bloomberg said the move adds pressure in a Chinese market where large model providers have repeatedly cut prices to win developers and enterprise customers. DeepSeek’s latest cuts show that battle is now reaching newly launched flagship models, not just older ones. (bloomberg.com) (reuters.com)