Tencent Cloud Notices > Notice Details
[LLM Service TokenHub][Agent Development Platform] Notice on Decommissioning and Upgrade of Tencent Cloud DeepSeek-V4-Flash 0731 Model
2026-09-30 14:41:52
With model version upgrades and iterations, LLM Service TokenHub, Agent Development Platform will retire the DeepSeek-V4-Flash 0731 model, and will provide automatic switching for eligible users. Details are as follows.
I. Retirement Time and Scope
Retirement Time：Effective from 00:00:00 on October 31, 2026 (Beijing Time).
Platforms involved：LLM Service TokenHub, Agent Development Platform.
Model to be retired: DeepSeek-V4-Flash 0731 GA (deepseek-v4-flash-0731).
II. Recommended Migration Plan
We recommend that you migrate to the new version of the DeepSeek series models (not directly provided by the original vendor) before the retirement, and update the model in your code or application configuration. Different models may differ in performance, pricing, and applicable scope. Please complete business verification in advance and confirm the relevant billing rules. For details, see Model Pricing and Enterprise Token Plan Credit Deduction Rules.
As of the publication of this announcement, the latest version of the DeepSeek series models is DeepSeek-V4.1-Flash (deepseek-v4.1-flash). For the latest available models, see Model List.
III. Automatic System Switching
1. Switching Time and Target
To ensure service continuity as much as possible, starting from 00:00:00 on October 31, 2026 (Beijing Time), the system will automatically switch users who have not completed migration and meet the eligibility conditions to the latest version of the DeepSeek series models (not directly provided by the original vendor) available on the platform at the time of the actual switch.
Between the publication of this announcement and the actual retirement, the platform may release newer versions. Therefore, the actual target model of the automatic switch may not be DeepSeek-V4.1-Flash (deepseek-v4.1-flash). Please stay tuned to the Model List.
2. Applicable Conditions
Automatic switching applies to either of the following types of users:
User Type | Automatic Switching Condition |
Postpaid API access users | Pay-as-you-go billing has been enabled for the target model of the automatic switch. For how to enable it, see Enablement Management |
Token Plan Enterprise Edition users | Users who call the model to be retired via Token Plan Enterprise Edition |
For users who do not meet the above conditions, the system will not perform automatic switching. After the model is retired, requests that continue to use the DeepSeek-V4-Flash 0731 model will fail. Please complete migration in advance.
3. Billing After Switching
Calls after the switch will be billed according to the billing rules of the target model, and the charges may differ from those of the original model. Before the switch, please confirm the pricing and applicable scope of the target model via the console or the relevant billing documentation.
4. Business Impact and Your Options
● Call impact: During the switch, request timeouts of a few seconds may occur. We recommend configuring appropriate timeout and retry mechanisms in advance.
● Performance differences: The output of the new and old models may differ. After the switch, if the new model does not meet your business needs, you can manually switch to other available models.
● Opting out of automatic switching: Please replace the model or stop the relevant calls on your own before the retirement. If you need assistance, please submit a ticket in advance to confirm the handling plan.
Thank you for your understanding and support. Tencent Cloud will continue to iterate its large model product portfolio to provide you with more powerful model capabilities and services. If you have any questions during use, please feel free to submit a ticket for feedback, and we will verify and handle them for you as soon as possible.
