DeepSeek V4 Pro API Integration: Model Names, Endpoints and Migration Checks

DeepSeek lists deepseek-v4-pro as an official API model name. For direct access, its documented OpenAI-compatible base URL is https://api.deepseek.com; its Anthropic-compatible base URL is https://api.deepseek.com/anthropic.

When configuring LLMAPI instead, treat the model request ID, protocol and billing mode as separate choices. Copy the exact request ID from your applicable LLMAPI model list or channel catalogue rather than assuming that the official DeepSeek name establishes account access.

Choose the provider and protocol first

Use the base URL for the service receiving your request and the protocol used by your client:

These are base URLs, not interchangeable complete request endpoints. DeepSeek's OpenAI-format HTTP example posts to https://api.deepseek.com/chat/completions. LLMAPI's client setup explicitly includes /v1 for OpenAI clients and omits it from the Anthropic SDK base URL.

Consult the DeepSeek API documentation for direct integration and the LLMAPI client setup documentation for LLMAPI configuration. The LLMAPI base URLs describe client setup; they do not establish that a particular model supports both protocols.

Make a direct OpenAI-compatible request

The following minimal curl example adapts DeepSeek's documented chat request structure to its listed deepseek-v4-pro model name. It is a direct DeepSeek request, not an LLMAPI request or a report of a tested response. Replace YOUR_API_KEY with your DeepSeek API key locally.

export DEEPSEEK_API_KEY='YOUR_API_KEY'

curl https://api.deepseek.com/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer ${DEEPSEEK_API_KEY}" \
  -d '{
    "model": "deepseek-v4-pro",
    "messages": [
      {"role": "system", "content": "You are a helpful assistant."},
      {"role": "user", "content": "Hello!"}
    ],
    "stream": false
  }'

DeepSeek also documents using the OpenAI SDK with an API key and a custom base_url, then calling client.chat.completions.create. The request above starts with a non-streaming conversation. DeepSeek documents setting stream to true to request streaming; validate that separately before making it part of your migration.

Keep this initial check focused on the chosen model, credentials and request format. Add application-specific options only after checking their documentation for your intended model and protocol.

Keep Anthropic configuration separate

DeepSeek describes its API as compatible with both OpenAI and Anthropic formats. For an Anthropic-compatible client connecting directly to DeepSeek, use its documented https://api.deepseek.com/anthropic base URL rather than the OpenAI base URL above.

Do not treat the curl example as an Anthropic request: it uses the OpenAI chat-completions format. Follow the Anthropic-format guidance linked from the DeepSeek documentation when configuring that client.

For LLMAPI, the Anthropic SDK base URL is instead https://llmapi.pro, with no /v1 suffix. Before choosing it for this model, check the selected catalogue entry's protocol support. General client compatibility and model-specific access are separate checks.

Verify the LLMAPI request ID and billing access

DeepSeek's official model name is a vendor specification. LLMAPI's catalogue determines which request ID to use with LLMAPI and which billing mode and protocol apply. Do not use a display name as a substitute for the request ID.

Use the LLMAPI model selection guide to choose the correct list:

• Monthly subscription: use a subscription key. Supported models share the plan quota, and the subscription model list is synchronized with GET /v1/models.
• Pay-as-you-go API: use an API billing key and its separate balance. Check the channel catalogue for model selection, supported protocols and current prices.

Before routing a V4 Pro request through LLMAPI, locate its entry in the list applicable to your account, copy the exact request ID and confirm the protocol. Do not assume inclusion in a subscription merely because your client supports OpenAI-format requests.

API credit does not increase subscription quota. Exhausting a subscription also does not automatically switch requests to API billing: choose the appropriate channel and key. For current plan details, use the subscription plans page; for metered model rates, use the channel catalogue.

Migration checklist

Before replacing an existing configuration, check each item independently:

1. Destination: are you calling DeepSeek directly or LLMAPI?
2. Protocol: does the client use OpenAI or Anthropic request format, and does the selected model entry support that format?
3. Base URL: does it match the service and protocol in the table above?
4. Model identifier: use deepseek-v4-pro for the documented direct DeepSeek model; for LLMAPI, copy the account catalogue's exact request ID.
5. Credentials and billing: select the key for the destination and, on LLMAPI, the intended subscription or API billing mode.
6. Initial request: validate a short non-streaming exchange before testing streaming or application-specific options.

Keep the initial migration limited to these configuration checks. A compatible API format is not a reason to assume identical model availability, subscription coverage or support for every option in an existing application.