Skip to main content
Doubao-Seed-2.1-Pro is ByteDance’s flagship model for the coding and agent era, available through Anyfast via an OpenAI-compatible interface. It is the strongest variant of the Seed 2.1 family — built for high-complexity work such as complex coding, long-horizon agents, and multi-step engineering delivery, with a 256K context window and further upgraded visual understanding.

Key capabilities

  • OpenAI-compatible — Works as a drop-in replacement with the OpenAI SDK
  • 256K context window — Handles project-scale context, with up to 256K tokens of output in a single response
  • Deep thinking — Chain-of-thought reasoning, on by default, with adjustable reasoning_effort
  • Strong coding & agents — Requirement understanding, long-horizon planning, continuous repair, and engineering delivery
  • Multimodal understanding — Text, image, video, and document understanding (audio not supported)
  • Function calling, structured output & context caching — Robust tool use, JSON output (beta), and prompt caching
  • Streaming — Real-time token streaming via SSE

Output specifications

PropertyValue
Input modalityText, Image, Video, Document
Output modalityText
Context window256K tokens
Max output tokens256K

Quick example

curl https://www.anyfast.ai/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "doubao-seed-2.1-pro",
    "messages": [
      { "role": "user", "content": "Explain quantum entanglement in simple terms." }
    ]
  }'

Thinking mode

Doubao-Seed-2.1-Pro reasons before answering by default. thinking.type is enabled unless you set it to disabled to skip reasoning for lightweight tasks. Use reasoning_effort to tune how long the model thinks.
Python
response = client.chat.completions.create(
    model="doubao-seed-2.1-pro",
    messages=[
        {"role": "user", "content": "Design a REST API for a blogging platform."}
    ],
    extra_body={
        "thinking": {"type": "enabled"},
        "reasoning_effort": "high"
    }
)

print(response.choices[0].message.content)
reasoning_effort controls reasoning length (effective only when thinking is enabled) and accepts minimal, low, medium, and high. Default: high.

Parameters

ParameterTypeRequiredDescription
modelstringYesMust be doubao-seed-2.1-pro
messagesarrayYesList of { role, content } objects
thinkingobjectNo{ "type": "enabled" | "disabled" }. Controls deep thinking. Default: enabled
reasoning_effortstringNominimal, low, medium, high. Effective when thinking is enabled. Default: high
max_tokensintegerNoMaximum tokens to generate (up to 262144)
temperaturefloatNo02. Controls randomness. Default: 1
top_pfloatNoNucleus sampling threshold. Default: 1
streambooleanNoEnable SSE streaming. Default: false
toolsarrayNoFunction tool definitions for tool use
response_formatobjectNo{ "type": "json_object" } for structured JSON output (beta)
stopstring / arrayNoSequences that stop generation

API Reference

View the interactive API playground for Doubao Seed 2.1 Pro.