Changes

Starting from April 29, 2026, 23:59:59 (Beijing Time), Model Studio will reduce the unit price for Implicit Cache billing of the DeepSeek-V4-Pro model. The specific adjustments are as follows:

Notes:

• Implicit caching takes effect only when a request hits the cache; input tokens that hit the cache will be billed at the cached_token rate.
• Input tokens that do not hit the cache will continue to be billed at the standard input_token rate.
• This adjustment applies solely to the implicit cache portion; the base inference pricing for the model remains unchanged.

Thank you for your support and trust in Model Studio.