SeaWhale AI

English

简体中文

English

简体中文

Appearance

Sign in

Doubao-Seed-2.1-lite

doubao-seed-2-1-lite-260915
Get API keyAPI reference

A new-generation omni-modal large model that natively supports unified understanding of text, vision, and audio, features a 1M ultra-long context window, and delivers excellent comprehensive multimodal capabilities. On 26/10/15, support for System One decision API capabilities was added.

Context window1.0M
Provider豆包
Released2026/10/20

Playground

Pricing

The same model is available through multiple service channels — choose based on latency, reliability and cost.

Prices in $ / 1M tokens
To pick a channel, add a provider field to the request body, for example "provider": { "channel": "direct" }. Valid values are direct / stable / economical; omit it to use the default channel.

Direct

Direct upstream connection — best when you need native behavior and the full context window.

InputOutputCache readCache write
0.12/M0.40/M0.02/M<0.01/M

Overview

Input
Text Image
Output
Text

Doubao-Seed-2.1-lite API: A Lightweight Tier with 1024K Context and 256K Output

Doubao-Seed-2.1-lite is the lightweight tier of ByteDance's Doubao Seed2.1 series. The Seed2.1 series first launched the Pro version on June 23, 2026, at the Volcano Engine FORCE Conference, focusing on agents, end-to-end coding, and multimodal understanding in real productivity scenarios; the Lite version then launched on Volcano Ark as version 260915 in September 2026 and was integrated into Doubao Work, handling high-frequency office tasks such as daily Q&A, document writing, spreadsheet processing, and PPT creation.

It is the "new tri-modal understanding model" on Volcano Ark, natively understanding text, vision, and audio. Compared with the previous generation Doubao-Seed-2.0-lite (the 260428 version's official specs are 256K context and 128K output), 2.1-lite's context window has expanded to 1024K, and both the maximum answer and maximum chain of thought have increased to 256K, fully matching the specs of the same-series Pro version 260915. When integrating it, Doubao Work highlighted two optimizations: faster responses and lower quota consumption.

Two things to know in advance: Deep thinking is enabled by default; to disable it during text generation, pass thinking.type=disabled; although the maximum answer is 256K, the default output is only 4K, so long-output tasks must explicitly set max_tokens.

SeaWhale AI provides Doubao-Seed-2.1-lite through an OpenAI-compatible API, supporting multimodal input, deep thinking, tool calling, and structured output.

Get API Key · Model ID: doubao-seed-2-1-lite-260915

Why Choose Doubao-Seed-2.1-lite

  • 1024K context — 4× the Doubao-Seed-2.0-lite 260428 version (256K), on par with 2.1-pro
  • 256K maximum answer — output limit is 2× the 2.0-lite 260428 version (128K)
  • Native understanding of text, vision, and audio — screenshots, videos, and speech do not need to be split across multiple specialized models
  • Pro-level specs at lightweight pricing — context, output, and chain-of-thought limits match 2.1-pro
  • Complete agent toolchain — function calling, MCP, knowledge base, web plugins, GUI task processing
  • Implicit + explicit context caching — significantly reduces cost for repeated calls with long prefixes

Core Capabilities

01 Ultra-long Context and Long Output

The 1024K context window is enough to hold a complete contract set, a medium-sized code repository, or a long meeting transcript in one go; the 256K maximum answer lets long reports, whole-chapter rewrites, and batch structured extraction complete in a single call, without slicing and stitching.

  • Maximum input 1024K, maximum chain of thought 256K
  • Default answer length 4K; expand to 256K via max_tokens as needed
  • Supports implicit caching, as well as prefix caching and Session caching under the Responses API

02 Tri-modal Understanding

2.1-lite natively understands text, images/videos, and audio, continuing the multimodal approach established in the May 2026 upgrade of 2.0-lite. When Doubao Work launched, it specifically noted stronger image, video, and audio understanding, with improved results in multimodal scenarios such as education and office work.

  • Understanding and Q&A for screenshots, charts, and document images
  • Video content understanding
  • Direct understanding of speech input (audio input is billed separately)

03 Agents and GUI Tasks

Volcano Ark classifies it under the "GUI task processing" capability category and supports a complete tool-calling ecosystem, making it suitable as a low-cost agent foundation.

  • Function Calling and structured output (the json_schema mode is recommended)
  • Built-in knowledge base, MCP, web content plugins, and image processing tools
  • Read interface screenshots and plan operation steps

04 High-frequency Office Productivity

Lite's core base is high-frequency, high-volume office traffic. Doubao Work uses it as one of the default lightweight models, covering daily Q&A, document writing, spreadsheet processing, and PPT creation, and highlights faster responses and lower quota consumption.

  • Document drafting, rewriting, and summarization
  • Spreadsheet data organization and analysis
  • Presentation outline and content generation

Best Use Cases

Scenario Description
Office assistant High-frequency tasks such as document writing, spreadsheet processing, and PPT generation
Long document processing Process contracts, tender documents, and long reports in one go with 1024K context
Long output generation 256K answer limit; whole-chapter rewrites and batch extraction completed in one call
Multimodal customer service Problem diagnosis after users send screenshots, voice, or videos
Lightweight agents Low-cost execution layer for tool calling, MCP, and GUI tasks
Batch structured extraction json_schema structured output for downstream systems

Differences Between Doubao-Seed-2.1-lite, Doubao-Seed-2.0-lite, and Doubao-Seed-2.1-pro

Capability Doubao-Seed-2.1-lite Doubao-Seed-2.0-lite Doubao-Seed-2.1-pro
Model ID doubao-seed-2-1-lite-260915 doubao-seed-2-0-lite-260215 doubao-seed-2-1-pro-260915
Release date September 2026 February 2026 (May upgrade to full modality) Series debut June 23, 2026
Context window 1024K tokens 1 million tokens (SeaWhale AI configuration) 1024K tokens
Maximum output 256K tokens 65.5K tokens (SeaWhale AI configuration) 256K tokens
Maximum chain of thought 256K tokens — 256K tokens
Input modalities Text, vision, audio Text, image, video, audio Text, vision
Deep thinking Enabled by default, can be disabled Enabled by default, can be disabled Enabled by default, can be disabled
Positioning Lightweight high-frequency, office and multimodal Balanced enterprise scenarios Series flagship, coding and agents

Specs come from the Volcano Ark model list; the Doubao-Seed-2.0-lite column reflects the current SeaWhale AI platform configuration. The official independent benchmark results for 2.1-lite have not yet been published. Specific billing is subject to the real-time price card at the top of the page.

FAQ

When was Doubao-Seed-2.1-lite released? The Seed2.1 series first launched the Pro version on June 23, 2026. The first Lite version, 260915, launched on Volcano Ark in September 2026 and was integrated into Doubao Work the same month.

What has been upgraded compared with Doubao-Seed-2.0-lite? In terms of specs, context has expanded from 256K (official value for 2.0-lite 260428) to 1024K, and the maximum answer has increased from 128K to 256K. In terms of capabilities, it is based on the Seed2.1 generation of training and has enhanced multimodal understanding; when Doubao Work launched, it also highlighted faster responses and lower quota consumption.

What are the context and output limits? Context and maximum input are both 1024K tokens, maximum answer is 256K tokens, and maximum chain of thought is 256K tokens. Note that the default answer length is only 4K; when long output is needed, explicitly set max_tokens.

How do I configure deep thinking? It is enabled by default. For the Chat API, pass thinking.type=disabled to disable it; for the Responses API, pass reasoning.effort=minimal. Some third-party developers report that it only accepts enabled / disabled and does not support the auto of 2.0-lite, so check this parameter when migrating from 2.0-lite.

Which tools and features are supported? Supports function calling, structured output (the json_schema mode is recommended), knowledge base, MCP, web content plugins, image processing tools, as well as implicit caching and explicit caching (prefix caching, Session caching).

Should I choose 2.1-lite or 2.1-pro? Both have the same context, output, and chain-of-thought limits. Choose Lite for high-frequency office work, multimodal understanding, and batch processing for lower cost and faster responses; choose Pro for the most complex coding, long-chain agents, and open-ended research tasks. The highest scores Seed2.1 achieved on benchmarks such as GDPval and MobileWorld were all from Pro.

Are there any shortcomings? Two things to note. First, the official evaluation results currently published are only for Seed2.1 Pro; Lite has no independent benchmark data, so before selection it is best to test with your own tasks. Second, when launching Seed2.1, the official team acknowledged that the series still has room for improvement on the most challenging open-ended tasks and frontier research questions, and the lightweight tier is even less advantaged in such tasks.

Why Choose SeaWhale AI to Use the Doubao-Seed-2.1-lite API

  • No Volcano Engine account needed — call directly with one key, no separate application for Ark model access
  • OpenAI-compatible API — multimodal request format is consistent with OpenAI; just change the model name in existing code
  • Compare old and new versions under the same account — share keys with Doubao-Seed-2.0-lite and 2.0-pro for smooth migration
  • RMB settlement — domestic models are priced in RMB

API

API integration ​

Model IDUse this value as the model in inference requests
doubao-seed-2-1-lite-260915
API KeyBearer token used to authenticate inference requests
Manage API keys
Base URLOpenAI compatible · /chat/completions
OpenAIhttps://api.haijingai.com/v2/
provider OptionalSelects a service channel; omit it and the system picks the default
"provider": { "channel": "direct" }

doubao-seed-2-1-lite-260915 usage examples ​

SeaWhale AI is compatible with the OpenAI API protocol, so you can call it with the OpenAI SDK or plain HTTP requests. Streaming is enabled by default.

js
curl https://api.haijingai.com/v2/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer <API_KEY>" \
  -d '{
    "model": "doubao-seed-2-1-lite-260915",
    "messages": [
      {"role": "system", "content": "You are a helpful assistant."},
      {"role": "user", "content": "Hello!"}
    ],
    "provider": { "channel": "direct" },
    "stream": true
  }'
# provider is optional — remove this line to use the default channel
js
from openai import OpenAI

client = OpenAI(
    base_url="https://api.haijingai.com/v2",
    api_key="<API_KEY>",
)

stream = client.chat.completions.create(
    model="doubao-seed-2-1-lite-260915",
    messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "Hello!"},
    ],
    stream=True,
    # Optional: pick a service channel; omit to use the default
    extra_body={"provider": {"channel": "direct"}},
)

for chunk in stream:
    if chunk.choices[0].delta.content:
        print(chunk.choices[0].delta.content, end="", flush=True)
js
import OpenAI from 'openai'

const client = new OpenAI({
  baseURL: 'https://api.haijingai.com/v2',
  apiKey: '<API_KEY>',
})

const stream = await client.chat.completions.create({
  model: 'doubao-seed-2-1-lite-260915',
  messages: [
    { role: 'system', content: 'You are a helpful assistant.' },
    { role: 'user', content: 'Hello!' },
  ],
  stream: true,
  // Optional: pick a service channel; omit to use the default
  // @ts-expect-error provider is a SeaWhale AI extension, not in the OpenAI SDK types
  provider: { channel: 'direct' },
})

for await (const chunk of stream) {
  process.stdout.write(chunk.choices[0]?.delta?.content ?? '')
}
Contact support