DeepSeek V4.1 Flash

DeepSeek V4.1 Flash is available through AnyFast with the public model ID deepseek-v4.1-flash and the OpenAI-compatible POST /v1/chat/completions endpoint. It supports text and image input, thinking and non-thinking modes, function calling, JSON output, streaming, and prompt caching.

DeepSeek calls this release DeepSeek-V4.1-Flash and uses deepseek-flash on its own API. When calling AnyFast, use deepseek-v4.1-flash.

Model specifications

Property	Value
AnyFast model ID	`deepseek-v4.1-flash`
Input	Text and images
Output	Text
Context window	1,048,576 tokens
Maximum output	393,216 tokens
Thinking mode	Supported and enabled by default
Thinking effort	`low`, `high`, `max`; default `high`
Structured output	JSON mode
Tools	Function calling
Streaming	SSE

Quick examples

Text
Vision
Thinking

cURL

curl https://www.anyfast.ai/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-v4.1-flash",
    "messages": [
      {"role": "user", "content": "Explain how a B-tree index accelerates database queries."}
    ],
    "max_tokens": 1024
  }'

Pass the user message as an array containing text and image_url blocks. The image can use a public HTTP(S) URL or a Base64 data URL.

cURL

curl https://www.anyfast.ai/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-v4.1-flash",
    "messages": [{
      "role": "user",
      "content": [
        {"type": "text", "text": "Describe this image and read any visible text."},
        {"type": "image_url", "image_url": {"url": "https://example.com/chart.png"}}
      ]
    }]
  }'

thinking controls whether reasoning is enabled. Set reasoning_effort separately at the top level.

Python

from openai import OpenAI

client = OpenAI(
    api_key="YOUR_API_KEY",
    base_url="https://www.anyfast.ai/v1",
)

response = client.chat.completions.create(
    model="deepseek-v4.1-flash",
    messages=[
        {"role": "user", "content": "Find and explain the bug in this algorithm."}
    ],
    reasoning_effort="high",
    extra_body={"thinking": {"type": "enabled"}},
)

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

Thinking mode

Thinking mode is enabled by default with high effort. Use {"thinking":{"type":"disabled"}} for latency-sensitive requests that do not need reasoning. When thinking is enabled, reasoning is returned in choices[0].message.reasoning_content and the final answer is returned in choices[0].message.content. The official model accepts low, high, and max effort. temperature, presence_penalty, and frequency_penalty have no effect in thinking mode. If a request includes tools, pass the complete reasoning_content from earlier assistant messages back in subsequent tool-call turns.

Image input

Use image_url blocks only in user messages. Supported provider formats are JPEG, PNG, GIF, and WebP. For URL inputs, use a directly downloadable HTTP(S) address; for local images, use a Base64 data URL. Image tokens are included in usage.prompt_tokens.

Common parameters

Parameter	Type	Required	Description
`model`	string	Yes	Must be `deepseek-v4.1-flash`.
`messages`	array	Yes	Conversation messages using `system`, `user`, `assistant`, or `tool` roles.
`thinking`	object	No	`{"type":"enabled"}` or `{"type":"disabled"}`. Thinking is enabled by default.
`reasoning_effort`	string	No	`none` disables thinking; `low`, `high`, and `max` select the effort. Default: `high`.
`max_tokens`	integer	No	Maximum generated tokens, from 1 through 393,216.
`stream`	boolean	No	Stream Chat Completions chunks over SSE.
`response_format`	object	No	Use `{"type":"json_object"}` for JSON mode and explicitly ask for JSON in the prompt.
`tools`	array	No	OpenAI-format function tools.
`tool_choice`	string or object	No	`none`, `auto`, `required`, or a named function.
`user_id`	string	No	Stable end-user identifier for content safety and cache isolation.