DeepSeek model names: the API IDs to know
DeepSeek’s official API documentation lists two model IDs: deepseek-flash and deepseek-v4-pro. It maps them to the version names DeepSeek-V4.1-Flash-1017 and DeepSeek-V4-Pro-0813, respectively. These version labels are not the API IDs, and the listed names should not be treated as a complete or permanent catalog. (api-docs.deepseek.com)
| API model ID | Version listed in the docs | Vision support listed in the docs |
|---|---|---|
deepseek-flash |
DeepSeek-V4.1-Flash-1017 | Supported |
deepseek-v4-pro |
DeepSeek-V4-Pro-0813 | Not supported |
The same table lists a 1M context length and a maximum output of 384K for both models. It also indicates that both support thinking and non-thinking modes. Check the live documentation for current details. (api-docs.deepseek.com)
Which name should you put in a production request?
For the Flash route, use deepseek-flash in the request’s model field. DeepSeek says this ID is served by DeepSeek-V4.1-Flash-1017. The legacy names deepseek-v4-flash and deepseek-v4-flash-vision-exp are still accepted, but their corresponding models are retired; requests using them are served by the Flash model and billed at the Flash price. An accepted legacy name does not mean its original model remains active. (api-docs.deepseek.com)
{
"model": "deepseek-flash"
}
For V4 Pro, the documented API ID is deepseek-v4-pro. DeepSeek also documents a List Models endpoint for checking currently available models and their metadata. (api-docs.deepseek.com)
Flash or V4 Pro? Choose by requirement, then test
- Need image input? The pricing table lists vision support for Flash and marks it unsupported for V4 Pro. (
api-docs.deepseek.com) - Comparing costs? The official page lists separate cached-input, uncached-input, and output rates. Flash’s listed rates are lower in each category. (
api-docs.deepseek.com) - Choosing for another workload? The listed shared features do not establish which model will perform better for your application; test candidates against representative tasks.
Listed API prices
The official pricing page states that model prices are shown per 1 million tokens and that usage billing is based on input and output tokens. These are the listed peak rates:
| Model ID | Cached input | Uncached input | Output |
|---|---|---|---|
deepseek-flash |
$0.006 | $0.30 | $1.20 |
deepseek-v4-pro |
$0.044 | $1.32 | $3.96 |
The page says off-peak rates are half the peak rates. Peak hours are 01:00–04:00 and 06:00–10:00 UTC, Monday through Friday, excluding Chinese public holidays. All other hours—including weekends and Chinese public holidays—are off-peak. DeepSeek notes that prices may change, so check the live pricing page before estimating deployment costs. (api-docs.deepseek.com)
Bottom line: use deepseek-flash for the documented Flash API route, including image-input use cases. Evaluate deepseek-v4-pro against your workload, noting that the pricing table lists vision as unsupported. Confirm current IDs, capabilities, and prices in DeepSeek’s documentation before deployment. (api-docs.deepseek.com)