DeepSeek / Model reference
DeepSeek models,
without the name confusion.
Match the API identifier to the model version, then check the input types and limits that apply to your integration.
| API name | Model version | Input | Context / max output |
|---|---|---|---|
deepseek-flash | DeepSeek-V4.1-Flash | Text and images | 1M tokens / 384K tokens |
deepseek-v4-pro | DeepSeek-V4-Pro-0813 | Text | 1M tokens / 384K tokens |
Context and maximum output are provider-stated limits. They do not describe how much a particular task needs or predict the quality of a response.
API names are not fixed version labels.
The identifier your application sends and the name of the underlying model can differ. For repeatable comparisons, record both, along with the provider, date, prompt, tools and request settings.
DeepSeek’s current reference maps deepseek-flash to V4.1-Flash and deepseek-v4-pro to V4-Pro-0813. The Flash model accepts image and text input; the Pro model listed here is text-only.
Choose by the task you actually have.
If you need image input
Start by checking support for your input format. V4.1-Flash lists image input support; review the official vision documentation and limits before integrating it.
If you need predictable costs
Compare input and output rates, your expected context length and when requests run. A model’s maximum context is a capacity limit, not a suggested default.
What about older model names?
Aliases and availability change. DeepSeek’s changelog explains updates and migration paths. Its current changelog confirms that V4-Pro API service continues after September 14, 2026; an earlier retirement announcement should not be treated as the latest status.
The official chat app and third-party hosts may expose a different model selection. Check the specific service you plan to use.
Sources: Model reference and pricing · Changelog · V4.1-Flash model card. Checked September 18, 2026.