DeepSeek API Adds New DeepSeek-V4-Pro-0813 Model
Nashnova编辑部
DeepSeek quietly listed DeepSeek-V4-Pro-0813 on its API pricing page, supporting both reasoning and non-reasoning modes — but no changelog or performance benchmarks have been released yet.
What exactly showed up?
A new entry — DeepSeek-V4-Pro-0813 — appeared on the "Models & Pricing" page of DeepSeek's API documentation.
The model supports both reasoning mode and non-reasoning mode. This means → developers can toggle between deep inference and fast response on a single API endpoint.
In plain terms = one model, two speeds: turn reasoning on when you need it to think hard, off when you need a quick answer.
Why hasn't DeepSeek said anything yet?
As of the time this was spotted, DeepSeek has published no changelog, no performance benchmarks, and no blog post.
This means → the specific improvement — whether it is stronger reasoning, a longer context window, or cost optimization — is entirely unknown.
This reflects a recent DeepSeek pattern: list the model quietly in the docs first, then publish a formal announcement later.
What does this signal for developers and the market?
The "V4-Pro" label suggests a higher-tier variant within the V4 family; 0813 likely marks a date tag (August 13).
Until official details drop, developers can only watch the API docs and pricing page — there is not enough information to evaluate a migration.
In plain terms = the signal is out, but the menu only shows the name, not the ingredients — waiting for the formal announcement is the safer call.
Content is for reference only, not financial advice.