First-call API configuration

One-sentence takeaway

Use deepseek-flash as the model name for the first call; the old aliases are still accepted, but requests are served by DeepSeek-V4.1-Flash and billed at Flash pricing. The official example uses an OpenAI-compatible format, thinking enabled, reasoning_effort set to high, and a non-streaming response. Official first-call documentation

Use cases

• Suitable tasks: Integrating DeepSeek-V4.1-Flash into OpenAI-compatible SDKs, software, scripts, or Agent tools. Official first-call documentation

• Unsuitable tasks: Not specified.

• Applicable model version: DeepSeek-V4.1-Flash, with deepseek-flash as its API model name. Official pricing and model details

• Applicable clients, Agents, or APIs: OpenAI API format, OpenAI SDK, cURL, Python, and Node.js; the documentation also states that the API is compatible with the Anthropic format, but this page does not provide Anthropic first-call code. Official first-call documentation

• Recommended reasoning level and parameters: The page's first-call example uses thinking: {"type": "enabled"}, reasoning_effort: "high", and stream: false; the page does not provide a uniform recommendation suitable for all tasks. Official first-call documentation

Ready-to-use content

The following are official call configurations and complete examples, rather than a complete prompt. Apply for an API key first, then provide it to the client through the DEEPSEEK_API_KEY environment variable.

Endpoint and model configuration

The old names deepseek-v4-flash and deepseek-v4-flash-vision-exp are still available, but the corresponding old models have been retired, so use deepseek-flash where possible. Official first-call documentation

cURL

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

Python (OpenAI SDK)

# Please install OpenAI SDK first: `pip3 install openai`
import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ.get('DEEPSEEK_API_KEY'),
    base_url="https://api.deepseek.com")

response = client.chat.completions.create(
    model="deepseek-flash",
    messages=[
        {"role": "system", "content": "You are a helpful assistant"},
        {"role": "user", "content": "Hello"},
    ],
    stream=False,
    reasoning_effort="high",
    extra_body={"thinking": {"type": "enabled"}}
)

print(response.choices[0].message.content)

Node.js (OpenAI SDK)

// Please install OpenAI SDK first: `npm install openai`

import OpenAI from "openai";

const openai = new OpenAI({
        baseURL: 'https://api.deepseek.com',
        apiKey: process.env.DEEPSEEK_API_KEY,
});

async function main() {
  const completion = await openai.chat.completions.create({
    messages: [{ role: "system", content: "You are a helpful assistant." }],
    model: "deepseek-flash",
    thinking: {"type": "enabled"},
    reasoning_effort: "high",
    stream: false,
  });

  console.log(completion.choices[0].message.content);
}

main();

Testing or workflow steps

1. Apply for an API key on the DeepSeek platform and set DEEPSEEK_API_KEY. API key entry

2. Use https://api.deepseek.com as the OpenAI base_url, and set the model name to deepseek-flash.

3. Install the corresponding OpenAI SDK for the client, then run the cURL, Python, or Node.js example above.

4. The example uses a non-streaming request; for a streaming response, change stream to true. Official first-call documentation

Original evidence and data

• The first-call page states that the API is compatible with the OpenAI and Anthropic formats, and lists the OpenAI endpoint https://api.deepseek.com, the Anthropic endpoint https://api.deepseek.com/anthropic, and the model name deepseek-flash.

• A model footnote states that deepseek-v4-flash and deepseek-v4-flash-vision-exp are still accepted, but requests are served by DeepSeek-V4.1-Flash and billed at Flash pricing. Official first-call documentation

• The model details page lists version DeepSeek-V4.1-Flash, a context length of 1M, and a maximum output of 384K, and states that both non-thinking and thinking modes are supported (thinking by default). Official pricing and model details

• The official first-call code includes thinking, reasoning_effort, and stream; the documentation explicitly says this is a non-streaming example and that stream=true provides a streaming response. Official first-call documentation

Scope and limitations

• deepseek-flash has a concurrency limit of 2,500 per account; concurrency is calculated per account and is unrelated to the API key. Requests exceeding the limit return HTTP 429. Official rate-limit and isolation documentation

• Non-streaming requests may continue to receive blank lines, while streaming requests may receive the SSE keep-alive comment : keep-alive; if inference has not started 10 minutes after the request is sent, the server closes the connection. Official rate-limit and isolation documentation

• The context and maximum output limits come from the model details page; the first-call page does not state other task-level limits for a single request. Official pricing and model details

• Although the old model names are still accepted, the old models have been retired; new integrations should consistently use deepseek-flash. Do not infer independent model capabilities or versions for the old names from this behavior. Official first-call documentation

Reproduction notes

On 2026-09-16, the DeepSeek official first-call, model details, and rate-limit pages were accessed and checked only through the Tabbit browser; no real API request was executed. The code, model names, endpoints, and limits in the document all come from text visible on those pages, and the API key uses an environment-variable placeholder.