DeepSeek V4.1 Pro release status
DeepSeek V4.1 Pro Release Status: V4 Pro API & Model IDs
- Author
- DeepSeekAgent.io Editorial Team
- Published
- Updated
As of September 28, 2026, DeepSeek has not released a model named “V4.1 Pro.” The official API currently exposes both deepseek-flash and deepseek-v4-pro: the former serves DeepSeek V4.1 Flash, while the latter continues to serve DeepSeek V4 Pro 0813. They are separate routes; deepseek-v4-pro has not been redirected to V4.1 Flash.
That distinction matters because the models differ in vision support, concurrency, pricing, and behavior. A V4.1 Flash release does not imply that the Pro route has also moved to V4.1.
Current official status
| API model ID | Current model | Image input | Context | Maximum output |
|---|---|---|---|---|
deepseek-flash | DeepSeek V4.1 Flash | Supported | 1M | 384K |
deepseek-v4-pro | DeepSeek V4 Pro 0813 | Not supported | 1M | 384K |
When V4.1 Flash launched on September 10, DeepSeek initially planned to discontinue the V4 Pro API on September 14. The company later changed that decision in response to user demand and kept V4 Pro available under its existing billing rules. The correct interpretation is two parallel model routes, not a temporary Pro-to-Flash alias.
Why the routes are easy to confuse
V4.1 Flash replaced the earlier Flash generation and retained compatibility mappings for some retired Flash IDs. V4 Pro follows a separate service path. Reading only the Flash migration notice can therefore lead to an incorrect conclusion about the Pro route.
To confirm a route, check the official changelog, the Models & Pricing page, and the /models response together. As long as deepseek-v4-pro is listed independently, it should not be treated as an alias for V4.1 Flash.
How to configure DeepSeek Harness
Choose the explicit model ID for the task:
- Use
deepseek-flashfor image input, higher concurrency, or lower cost. - Use
deepseek-v4-proto continue established V4 Pro text workloads and preserve an existing evaluation baseline. - Do not treat
deepseek-v4-proas an abstract “latest Pro” alias. It currently identifies V4 Pro 0813.
After changing a Profile, run a fixed task set and record the model ID, endpoint, DSH version, runtime, tokens, tool turns, and result quality. This makes a future route change distinguishable from a Harness configuration or task change.
What would count as a V4.1 Pro release
A release is confirmed only when DeepSeek publishes the model name together with an API route or mapping, pricing, and model documentation. Community discussion, placeholder labels, and predictions about the next version are not release evidence.
Until that happens, interfaces and examples should label deepseek-v4-pro as V4 Pro 0813 and deepseek-flash as V4.1 Flash.
Related reading
- DeepSeek V4 Pro vs V4.1 Flash
- DeepSeek V4.1 Flash overview
- Configure V4.1 Flash in DeepSeek Harness
- V4.1 Flash vs V4 Flash in Harness workloads
References
- DeepSeek API updates
- DeepSeek Models & Pricing
- DeepSeek API: List Models
- DeepSeek V4 Pro general availability
FAQ
Has DeepSeek released V4.1 Pro?
No. As of September 28, 2026, the official public models remain V4.1 Flash and V4 Pro 0813; there is no formal V4.1 Pro model or API route.
Does deepseek-v4-pro now return V4.1 Flash?
No. Official documentation lists it separately as DeepSeek V4 Pro 0813. V4.1 Flash uses deepseek-flash.
Which model should a new project choose?
Start with deepseek-flash for vision, high-concurrency, and cost-sensitive workloads. Use deepseek-v4-pro when an existing text workflow depends on the V4 Pro baseline. Make the final choice using quality, latency, and cost on the same task set.