Passer au contenu principal
AITier

Seed 2.1 Turbo

ByteDanceProprietary

Description

ByteDance's low-cost, low-latency next-generation agent model for large-scale production. A deep-thinking model with the full feature set of Seed 2.1 Pro and comparable performance, it is designed for enterprise-grade deployments handling high volumes of online calls across coding, long-chain agents, and multimodal tasks. Served via Volcano Engine as Doubao-Seed-2.1-turbo.

Date de sortie
2026-06-24
Paramètres
—
Longueur du contexte
262K
Modalités
image, text, video

Radar de capacités

60
general
50
coding
60
reasoning
51
scienceest.
70
agents
70
multimodal

Science est estimé à partir des scores scientifiques de LLM Stats ou du raisonnement lorsque les benchmarks scientifiques dédiés ne sont pas disponibles.

Classements

Domaine#RangScoreSource
Capacité agentique61
50.0
LS
Classement multimodal25
58.0
LS

Scores de benchmarks (LLM Stats)

(LLM Stats (zeroeval))

3d

BLINK79.4%Aut.

Agents

BrowseCompOpenAI (2025)84.9%Aut.
GDPval82.7%Aut.
MCP Atlas80.3%Aut.
OSWorld76.4%Aut.
Web Bench73.6%Aut.
OfficeQA Pro71.1%Aut.
MobileWorld70.0%Aut.
Terminal-Bench 2.167.6%Aut.
CyberGym67.0%Aut.
OneMillion Bench66.6%Aut.
SeedClawBench63.8%Aut.
WildClawBench62.8%Aut.
Trae Code Gen59.7%Aut.
SWE-Bench ProPrinceton NLP (2024)57.0%Aut.
Trae Error Fix56.7%Aut.
xDailyBench56.4%Aut.
Finance Agent v1.156.0%Aut.
Workspace Bench54.7%Aut.
Agent Startup Bench54.0%Aut.
Program Bench49.4%Aut.
Toolathlon49.1%Aut.
Doubao Multi-Turn Bench49.0%Aut.
PresentBench48.3%Aut.
Repo Env46.7%Aut.
ClawEval-MM46.0%Aut.
NL2Repo43.7%Aut.
CreativeWork34.5%Aut.
SWE-Atlas30.6%Aut.
APEX-Agents29.2%Aut.
GameWorld25.9%Aut.
DeepSWE23.0%Aut.
PostTrainBench18.3%Aut.

Biology

SciCode57.8%Aut.

Chemistry

SuperGPQA67.4%Aut.
SuperChem56.6%Aut.

Code

FrontierCS50.8%Aut.
Artifacts Bench47.0%Aut.
Image2FloorPlan35.9%Aut.

Coding

AetherCode67.9%Aut.

Embodied

EmbSpatialBench0.82 / 100Aut.

General

MMMU-Pro82.2%Aut.
SimpleVQA0.71 / 100Aut.
KINA46.6%Aut.
MSQA42.0%Aut.

Image To Text

OCRBench_V262.8%Aut.

Knowledge

VideoSimpleQA71.4%Aut.
WorldBench63.7%Aut.

Long Context

DUDE83.1%Aut.
LongVideoBench80.6%Aut.
MMLongBench-128K76.9%Aut.
LVBench76.8%Aut.

Math

MathVision92.7%Aut.
MathVista90.5%Aut.
MathVerse89.2%Aut.
Beyond AIME88.0%Aut.
EMMA78.4%Aut.
FrontierScience Olympiad76.0%Aut.
DynaMath68.1%Aut.
Humanity's Last Exam54.6%Aut.
MathArena Apex35.4%Aut.
LiveMathematicianBench27.7%Aut.
HorizonMath2.0%Aut.

Multimodal

CharXiv-D94.6%Aut.
Video-MME89.0%Aut.
CharXiv-R83.6%Aut.
OVOBench79.2%Aut.
TVBench77.2%Aut.
LiveSports-3K77.1%Aut.
MotionBench74.8%Aut.
TreeBench71.1%Aut.
ChartQAPro70.9%Aut.
OVBench69.7%Aut.
VLMsAreBiased68.3%Aut.
VideoHolmes67.6%Aut.
Minerva65.9%Aut.
CrossVid63.2%Aut.
BabyVision62.9%Aut.
ContPhy61.1%Aut.
MeasureBench58.9%Aut.
TOMATO56.8%Aut.
VisuLogic0.53 / 100Aut.
WorldVQA48.6%Aut.
VisFactor43.9%Aut.
MMSIBench31.4%Aut.
ZEROBench0.11 / 100Aut.

Reasoning

ERQA71.3%Aut.
ArcAGI261.3%Aut.
FrontierScience Research33.3%Aut.

Spatial Reasoning

RealWorldQA86.3%Aut.

Indices d'évaluation AA

(Artificial Analysis)

Aucune donnée d'évaluation AA disponible

Scores par catégorie LLM Stats

(LLM Stats (zeroeval))
Structured Output
90
Legal
80
Search
80
Spatial Reasoning
80
Embodied
80
Finance
80
3d
80
Long Context
70
Multimodal
70
Safety
70
Image To Text
70
Healthcare
70
Economics
70
Tool Calling
70
Video
70
Vision
70
Math
60
Physics
60
Reasoning
60
Knowledge
60
General
60
Agents
60
Biology
60
Chemistry
60
Coding
60
Science
50
Frontend Development
50
Code
50
Systems
20

Tarification

Prix d'entrée$0.5 / 1M tokens
Prix de sortie$2.5 / 1M tokens
Prix mixte (3:1)$1 / 1M tokens

Vitesse

Aucune donnée de vitesse disponible

Classement des Prix par Fournisseur

Classement des Prix par Fournisseur

4 fournisseurs

Moins cher: ByteDancePlus cher: Kilo Gateway
FournisseurEntréeSortie
1ByteDancePRINCIPAL
$0.5
$2.5
2NanoGPT
$0.5
$2.5
3OpenRouter
$0.5
$2.5
4Kilo Gateway
$0.5
$2.5

Comparer les prix entre différents fournisseurs API pour ce modèle.

Sources externes

LLM StatsArtificial Analysis