Doubao-Seed-2.1-Pro (豆包旗舰) Official Specifications, Pricing & Benchmarks
ByteDance Seed 2.1 Pro multimodal flagship with 1024K context, 256K output, and low-latency China access.
📋 Doubao-Seed-2.1-Pro (豆包旗舰) Official API & System Card Specifications
Verified against official ByteDance 字节豆包 developer documentation & pricing
1. Identity & Capabilities
API Spec- API Model ID
- doubao-seed-2-1-pro-260628
- Release Date
- 2026-06-23
- Weight License
- Proprietary (Undisclosed)
- Native Reasoning
- ✅ Supported
- Function Calling & JSON Schema
- ✅ 原生支持
- Batch API 异步批处理
- ✅ Supported (50% off)
2. Context, Output & Modalities
Limits- Max Context Window
- 1,024,000 tokens (1024K)
- Max Output Tokens
- 256,000 tokens
- Input Modalities
- Text + Image + Video
- Output Modalities
- Text
- Direct China Access
- ✅ Direct CN + RMB
- Deployment Mode
- Cloud API only (Proprietary)
3. Official Pricing & Cache
CNY / 1M- Input / 1M Tokens
- ¥ 6
- Cached Input / 1M
- ¥ 1.2 -80%
- Output / 1M Tokens
- ¥ 30
- Tiered Pricing Notes
- Starts at ¥0.8/¥2.0 within 32K; tiered up to 1024K context & 256K max output
🏆 Official Benchmark Results
Sourced from official System Card / Model Card & technical reports
A more robust and challenging multi-task language understanding benchmark that extends MMLU by expanding multiple-choice options from 4 to 10, eliminating trivial questions, and focusing on reasoning-intensive tasks.
PhD-level Google-proof science reasoning questions
All 30 problems from the 2026 American Invitational Mathematics Examination (AIME I and AIME II), testing olympiad-level mathematical reasoning with integer answers from 000-999.
A verified subset of 500 software engineering problems from real GitHub issues, validated by human annotators for evaluating language models' ability to resolve real-world coding issues by generating patches for Python codebases.
A more robust multi-discipline multimodal understanding benchmark that enhances MMMU through a three-step process: filtering text-only answerable questions, augmenting candidate options, and introducing vision-only input settings.
🎛️ Official Sampling & Reasoning Parameters
Recommended generation settings & API parameter constraints
💡 Supports adaptive thinking toggle (enabled / disabled / auto) in Volcengine Ark API.
📊 Doubao-Seed-2.1-Pro (豆包旗舰) Full Benchmark Scorecard
70 public benchmarks · ranks among 154 models · synced 2026-11-20
💻 Coding 13 items
MCP Atlas is a benchmark for evaluating AI models on scaled tool use capabilities, measuring how well models can coordinate and utilize multiple tools across complex multi-step tasks.
Terminal-Bench 2.1 is an updated release of the Terminal-Bench benchmark that tests AI agents' ability to operate a computer via the terminal.
CyberGym is a benchmark for evaluating AI agents on cybersecurity tasks, testing their ability to identify vulnerabilities, perform security analysis, and complete security-related challenges in a controlled environment.
AetherCode is a competitive-programming benchmark of olympiad-level algorithmic coding problems.
WildClawBench is an agentic coding benchmark from InternLM/Claw-Eval that reports overall model performance on real-world tool-using development tasks.
SciCode is a research coding benchmark curated by scientists that challenges language models to code solutions for scientific problems.
SWE-Bench Pro is an advanced version of SWE-Bench that evaluates language models on complex, real-world software engineering tasks requiring extended reasoning and multi-step problem solving.
Artifacts Bench evaluates a model's ability to generate visual code artifacts, measuring the quality of generated interactive and visual front-end outputs from natural-language requests.
Program Bench evaluates code-generation agents by asking them to recreate a program's behavior from only a compiled binary and documentation.
NL2Repo evaluates long-horizon coding capabilities including repository-level understanding, where models must generate or modify code across entire repositories from natural language specifications.
SWE-Atlas is a software engineering benchmark focused on debugging, evaluating a model's ability to localize and fix bugs in real-world codebases.
DeepSWE is a software engineering agent benchmark evaluated with the mini-swe-agent harness, where each task is solved in an isolated container with no internet access. It measures an agent's ability to autonomously resolve real-world coding issues end to end.
PostTrainBench evaluates a model's ability to autonomously post-train base models. Given pretrain-only base models, the agent must complete the full pipeline of data synthesis, training, evaluation, and iteration within a time budget, scored across downstream benchmarks such as…
🤖 Agents & Tools 12 items
GDPval is an OpenAI benchmark evaluating AI models on economically valuable, real-world knowledge-work tasks spanning many professional occupations and industries.
BrowseComp is a benchmark comprising 1,266 questions that challenge AI agents to persistently navigate the internet in search of hard-to-find, entangled information.
OSWorld: The first-of-its-kind scalable, real computer environment for multimodal agents, supporting task setup, execution-based evaluation, and interactive learning across Ubuntu, Windows, and macOS with 369 computer tasks involving real web and desktop applications, OS file…
MobileWorld is a benchmark for evaluating multimodal agents on real mobile-device tasks, testing GUI grounding, navigation, and multi-step task completion in mobile environments.
OfficeQA Pro evaluates AI models on professional knowledge-work questions and tasks drawn from real office workflows, including document analysis, spreadsheet reasoning, and information synthesis across business domains.
OneMillion Bench evaluates AI agents on high-economic-value tasks that require sustained, reliable execution across long-horizon real-world workflows.
Finance Agent v1.1 is an agentic financial-analysis benchmark that evaluates models on real-world finance workflows, including retrieving and reasoning over financial documents and performing multi-step calculations.
Workspace Bench evaluates AI agents on high-economic-value workplace tasks that span multi-step planning, file processing, and tool use across realistic office and productivity workflows.
ClawEval-MM is the multimodal variant of ClawEval, evaluating agentic problem solving with visual inputs.
Tool Decathlon is a comprehensive benchmark for evaluating AI agents' ability to use multiple tools across diverse task categories. It measures proficiency in tool selection, sequencing, and execution across ten different tool-use scenarios.
Agents' Last Exam is a challenging benchmark for AI agents on hard, long-horizon tasks that test sustained reasoning, planning, and tool use, reported with and without tool access.
APEX-Agents is a benchmark evaluating AI agents on long horizon professional tasks that require sustained reasoning, planning, and execution across complex multi-step workflows.
📐 Math 10 items
MATH-Vision is a dataset designed to measure multimodal mathematical reasoning capabilities.
MathVista evaluates mathematical reasoning of foundation models in visual contexts. It consists of 6,141 examples derived from 28 existing multimodal datasets and 3 newly created datasets (IQTest, FunctionQA, and PaperQA), combining challenges from diverse mathematical and…
Beyond AIME is a difficult mathematical reasoning benchmark designed to test deeper reasoning chains and harder decomposition than standard AIME-style problem sets.
EMMA (Enhanced MultiModal reAsoning) is a benchmark for organic multimodal reasoning across mathematics, physics, chemistry, and coding.
FrontierScience Olympiad is a benchmark of olympiad-level scientific reasoning problems, testing expert understanding and multi-step reasoning across advanced natural-science domains.
A multimodal mathematics and reasoning benchmark focused on dynamic visual problem solving.
SuperGPQA is a comprehensive benchmark that evaluates large language models across 285 graduate-level academic disciplines.
Humanity's Last Exam (HLE) is a multi-modal academic benchmark with 2,500 questions across mathematics, humanities, and natural sciences, designed to test LLM capabilities at the frontier of human knowledge with unambiguous, verifiable solutions.
MathArena Apex is a challenging math contest benchmark featuring the most difficult mathematical problems designed to test advanced reasoning and problem-solving abilities of AI models. It focuses on olympiad-level mathematics and complex multi-step mathematical reasoning.
HorizonMath is an extremely difficult frontier mathematics benchmark designed to test the limits of mathematical reasoning on research-level and competition-beyond problems.
👁️ Multimodal 32 items
CharXiv-D is the descriptive questions subset of the CharXiv benchmark, designed to assess multimodal large language models' ability to extract basic information from scientific charts.
Video-MME is the first-ever comprehensive evaluation benchmark of Multi-modal Large Language Models (MLLMs) in video analysis.
RealWorldQA is a benchmark designed to evaluate basic real-world spatial understanding capabilities of multimodal models.
CharXiv-R is the reasoning component of the CharXiv benchmark, focusing on complex reasoning questions that require synthesizing information across visual chart elements.
VLMsAreBiased evaluates whether vision-language models rely on visual evidence or fall back on language priors when answering.
DUDE (Document Understanding Dataset and Evaluation) tests multi-page, multi-domain document understanding and reasoning.
A more robust multi-discipline multimodal understanding benchmark that enhances MMMU through a three-step process: filtering text-only answerable questions, augmenting candidate options, and introducing vision-only input settings.
BLINK: Multimodal Large Language Models Can See but Not Perceive. A benchmark for multimodal language models focusing on core visual perception abilities.
OVOBench (Online Video Online Benchmark) evaluates streaming video understanding, testing a model's ability to perceive and respond to video content in real time as it unfolds.
LongVideoBench is a question-answering benchmark featuring video-language interleaved inputs up to an hour long.
TVBench is a temporal video understanding benchmark evaluating reasoning over actions, events, and temporal dynamics in videos.
TOMATO (Temporal Reasoning Multimodal Evaluation) assesses multimodal models on motion and temporal perception in video, testing understanding of actions, motion, and changes over time.
LVBench is an extreme long video understanding benchmark designed to evaluate multimodal models on videos up to two hours in duration. It contains 6 major categories and 21 subcategories, with videos averaging five times longer than existing datasets.
LiveSports-3K evaluates fine-grained understanding and commentary of live sports video.
VideoSimpleQA evaluates factual knowledge grounded in video content, measuring how accurately models answer short, fact-seeking questions about videos.
MotionBench is a benchmark for evaluating multimodal models on motion understanding in videos, testing the ability to comprehend temporal dynamics, movement patterns, and action sequences.
A benchmark for early-stage visual reasoning and perception on child-like vision tasks.
Embodied Reasoning Question Answering benchmark consisting of 400 multiple-choice visual questions across spatial reasoning, trajectory reasoning, action reasoning, state estimation, and multi-view reasoning for evaluating AI capabilities in physical world interactions.
TreeBench evaluates visual grounded reasoning, requiring models to localize and reason about fine-grained visual details.
ChartQAPro is a challenging benchmark for question answering over diverse, real-world charts and infographics.
Minerva is a benchmark for complex video reasoning, evaluating models on multi-step reasoning over long and information-dense video content.
OVBench is an online video understanding benchmark that evaluates a model's ability to perceive, memorize, and reason about real-time video streams as they unfold.
VideoHolmes evaluates video understanding and reasoning capabilities in multimodal models.
CrossVid evaluates cross-video reasoning, requiring models to integrate information across multiple videos.
ContPhy is a continuum physical-reasoning benchmark evaluating understanding of physical dynamics in video.
OCRBench v2: Enhanced large-scale bilingual benchmark for evaluating Large Multimodal Models on visual text localization and reasoning with 10,000 human-verified question-answering pairs across 8 core OCR capabilities.
WorldVQA is a benchmark designed to evaluate atomic vision-centric world knowledge. It assesses models' ability to understand and reason about visual elements representing real-world knowledge.
VisFactor is a benchmark evaluating fine-grained visual factor perception and reasoning over images.
EmbSpatialBench evaluates embodied spatial understanding and reasoning capabilities.
SimpleVQA is a visual question answering benchmark focused on simple queries.
VisuLogic evaluates logical reasoning capabilities in visual contexts.
ZEROBench is a challenging vision benchmark designed to test models on zero-shot visual understanding tasks.
🧠 Reasoning & Science 3 items
ARC-AGI-2 is the second-generation Abstraction and Reasoning Corpus benchmark measuring fluid, general reasoning and abstraction.
SuperChem is a benchmark of advanced chemistry problems requiring expert-level domain knowledge and reasoning.
FrontierScience Research is a benchmark evaluating AI models on cutting-edge scientific research questions requiring deep domain expertise, multi-step reasoning, and synthesis of complex scientific concepts across disciplines.
Source: ZeroEval / llm-stats.com (✓ = third-party verified, others vendor-reported).
Closest Self-Hostable Open-Weight Alternatives to Doubao-Seed-2.1-Pro (豆包旗舰)
Open-weight models with full HuggingFace weights for private GPU / Mac deployment
Frequently Asked Questions
Is Doubao-Seed-2.1-Pro (豆包旗舰) open-source or proprietary?
Doubao-Seed-2.1-Pro (豆包旗舰) (doubao-seed-2-1-pro-260628) is a proprietary commercial model by ByteDance 字节豆包. Weights and internal layer dimensions are undisclosed; it is accessed via cloud API.
What are the context limit and API pricing of Doubao-Seed-2.1-Pro (豆包旗舰)?
Doubao-Seed-2.1-Pro (豆包旗舰) supports 1024K input context and 256,000 max output tokens, priced at ¥6/1M input (¥1.2/1M cached) and ¥30/1M output.