According to a research institution, a version of the flagship AI model under the Chinese startup DeepSeek was far less expensive to run than the world-famous model in benchmark tests, and was more than 100 times less expensive to run than Anthropic's Claude Fable 5. According to sources, DeepSeek is preparing for a potential initial public offering, and the company officially unveiled its V4-Flash model on Friday, the company's latest attempt to revive momentum by leveraging what it does best — providing an ultra-low-cost AI alternative. According to research firm Artificial Analysis, DeepSeek's V4-Flash model charges $0.14 per million input tokens and $0.28 per million output tokens. A token is a unit of data used to measure AI usage. San Francisco-based Artificial Analysis estimates that the average test cost for V4-Flash is 3 cents each time, while Moonshot AI's Kimi K3 is 86 cents, OpenAI's GPT-5.6 Sol is $1.86, and Claude Fable 5 is $3.15.

According to a research institution, a version of the flagship AI model under the Chinese startup DeepSeek was far less expensive to run than the world-famous model in benchmark tests, and was more than 100 times less expensive to run than Anthropic's Claude Fable 5. According to sources, DeepSeek is preparing for a potential initial public offering, and the company officially unveiled its V4-Flash model on Friday, the company's latest attempt to revive momentum by leveraging what it does best — providing an ultra-low-cost AI alternative. According to research firm Artificial Analysis, DeepSeek's V4-Flash model charges $0.14 per million input tokens and $0.28 per million output tokens. A token is a unit of data used to measure AI usage. San Francisco-based Artificial Analysis estimates that the average test cost for V4-Flash is 3 cents each time, while Moonshot AI's Kimi K3 is 86 cents, OpenAI's GPT-5.6 Sol is $1.86, and Claude Fable 5 is $3.15.