As the model is upgraded and iterated, LLM Service TokenHub will decommission the DeepSeek-V4-Flash series (Vendor Direct) models. After that, access to the above models will no longer be provided. To ensure the stable operation of your business, please pay close attention to the following guidance:
I. Decommissioning Time and Scope
Decommissioning time: Starting at 00:00:00 (Beijing Time) on December 10, 2026.
Decommissioned models (2 models, 4 model parameter values):
|
DeepSeek-V4-Flash (Vendor Direct) | deepseek-v4-flash-202605, deepseek/deepseek-v4-flash-0731, deepseek/deepseek-v4-flash |
DeepSeek-V4-Flash-Vision-Exp (Vendor Direct) | deepseek/deepseek-v4-flash-vision-exp |
II. Recommended Migration Plan
Platform Automatic Switch Instructions
1. Switch Time and Target
To help ensure business continuity, starting at 00:00:00 (Beijing Time) on December 10, 2026, the platform will automatically switch eligible users who have not completed migration to the DeepSeek-V4.1-Flash-1017 (Vendor Direct) model.
2. Eligibility Criteria:
|
Postpaid API users | The new model for the automatic switch already has pay-as-you-go billing enabled. For how to enable it, see Activation |
Token Plan subscribers, including Personal and Enterprise tiers | Using Token Plan API Key to call the models being decommissioned this time |
For users who do not meet the above conditions, the platform will not perform an automatic switch. After the models are decommissioned, continued requests to the decommissioned models will result in call errors. Please be sure to complete the migration in advance.
3. Billing After Switch:
After the switch, calls will be billed according to the billing rules of the new model, and the fees may differ from those of the original model. Before switching, please confirm the pricing and applicable scope of the new model via the console or the relevant billing documentation.
4. Business Impact and Options:
Call impact: Requests may experience timeouts of several seconds during the switch. We recommend configuring reasonable timeout and retry mechanisms in advance.
Performance consistency: The new and old models deliver the same output. If you have other business needs, you may manually switch to another available model.
If you do not want to use the automatic switch: Please complete the model replacement or stop related calls on your own before the decommissioning date. If you need assistance, please submit a ticket in advance to confirm the resolution plan.
Thank you for your understanding and support. Tencent Cloud will continue to iterate on its large model product portfolio to provide you with stronger model capabilities and services. If you encounter any issues during use, please submit a ticket for feedback. We will verify and resolve your issue as soon as possible.