Tencent Cloud DeepSeek-V3.2 will be offline on July 16

Tencent Cloud announced that the DeepSeek-V3.2 model of the original Knowledge Engine Atomic Capability Platform and large model service platform TokenHub will be officially offline starting from 00:00 on July 16, 2026, Beijing time, and access services to the model will no longer be provided at that time.

Officials recommend that users migrate to the DeepSeek-V4 series models with better capabilities. If the user does not complete the migration before the deadline, the system will automatically switch to the latest DeepSeek model to ensure uninterrupted service. In conjunction with this offline, Tencent Cloud Intelligent Development Platform has significantly reduced the price of the DeepSeek-V4 series models starting from June 3 - V4-Pro inference input and output prices have dropped by 75%, cache hit prices have dropped by as much as 97.5%; V4-Flash cache hit prices have dropped by as much as 90%.

The offline of V3.2 is the normal rhythm of model generation replacement, marking that the V4 series has completely replaced V3 as the main DeepSeek model on Tencent Cloud. Tencent Cloud chose to promote the retirement of V3.2 after a significant price reduction of the V4 series. The timing was just right - the price reduction reduced users' migration resistance, and the automatic switching mechanism provided a safety net. For enterprise users who are still using V3.2, it is imperative to complete the migration test before July 16: although the automatic switching will not cause service interruption, the V4 series may be different from V3.2 in prompt compatibility and output style. It is recommended to verify it in the test environment in advance.