Venice Atlas
← Back to model catalog

VENICE ATLAS · MODEL DOSSIER · SEED 2.1 TURBO

Model overview

Seed 2.1 Turbo

Seed 2.1 Turbo (Dola-Seed-2.1) is ByteDance’s next-generation multimodal model for the coding and agent era, with engineering-grade code delivery, long-horizon agent execution, and upgraded GUI and video understanding. Supports text, image, and video inputs with a 256K context window.

seed-2-1-turbo · venice.ai
Text · Venice catalog$0.63 per 1M tokensPrice on Venice AI
ModalityTextOn Venice AI
Primary price$0.63 per 1M tokensper 1M tokens
Context window256KMax tokens
Benchmarks90Benchmark results

Benchmark results

See the tests behind each score. They use the benchmark’s original scoring scale.

Biology#3030.10 ratingChemistry#1934.84 ratingEconomics#522.55 ratingVision#3130.36 ratingFinance#1030.71 ratingHealthcare#1930.26 ratingImage To Text#422.76 ratingKnowledge#812.58 ratingLegal#732.35 ratingLong Context#528.94 ratingMath#1339.32 ratingMultimodal#1832.99 ratingPhysics#2333.80 ratingReasoning#3744.19 ratingSafety#21-1.12 ratingScience#717.34 ratingSearch#1224.42 ratingSpatial Reasoning#1226.59 ratingStructured Output#1718.48 ratingSystems#102.25 ratingVideo understanding#226.51 rating
Agents19 results
Agent Startup Benchagent-startup-bench · agents
54.0%
ReportedOct 2026
ClawEval-MMclaw-eval-mm · agents
46.0%
ReportedOct 2026
CreativeWorkcreativework · agents
34.5%
ReportedOct 2026
Doubao Multi-Turn Benchdoubao-multi-turn-bench · agents
49.0%
ReportedOct 2026
Finance Agent v1.1finance-agent-v1.1 · agents
56.0%
ReportedOct 2026
GameWorldgameworld · agents
25.9%
ReportedOct 2026
MobileWorldmobileworld · agents
70.0%
ReportedOct 2026
OfficeQA Proofficeqa-pro · agents
71.1%
ReportedOct 2026
OneMillion Benchonemillion-bench · agents
66.6%
ReportedOct 2026
PostTrainBenchposttrainbench · agents
18.3%
ReportedOct 2026
PresentBenchpresentbench · agents
48.3%
ReportedOct 2026
Program Benchprogram-bench · agents
49.4%
ReportedOct 2026
SeedClawBenchseedclawbench · agents
63.8%
ReportedOct 2026
SWE-Atlasswe-atlas · agents
30.6%
ReportedOct 2026
Toolathlontoolathlon · agents
49.1%
ReportedOct 2026
Web Benchweb-bench · agents
73.6%
ReportedOct 2026
WildClawBenchwildclawbench · agents
62.8%
ReportedOct 2026
Workspace Benchworkspace-bench · agents
54.7%
ReportedOct 2026
xDailyBenchxdailybench · agents
56.4%
ReportedOct 2026
Chemistry1 result
SuperChemsuperchem · chemistry
56.6%
ReportedOct 2026
Coding8 results
Artifacts Benchartifacts-bench · code
47.0%
ReportedOct 2026
CyberGymcybergym · code
67.0%
ReportedOct 2026
DeepSWEdeepswe · code
23.0%
ReportedOct 2026
NL2Reponl2repo · code
43.7%
ReportedOct 2026
AetherCodeaethercode · coding
67.9%
ReportedOct 2026
Repo Envrepo-env · coding
46.7%
ReportedOct 2026
Trae Code Gentrae-code-gen · coding
59.7%
ReportedOct 2026
Trae Error Fixtrae-error-fix · coding
56.7%
ReportedOct 2026
General intelligence1 result
GDPvalgdpval · general
82.7%
ReportedAug 2026
Knowledge1 result
KINAkina · knowledge
46.6%
ReportedOct 2026

Learn what each benchmark measures ↗