Model

Doubao-1.5-Pro

CLOSED

ByteDance · Doubao family · released Jan 22, 2025

MoE flagship matching GPT-4o/Claude 3.5 Sonnet at far lower cost; adds a Deep Thinking mode that beats o1-preview on AIME

ReasoningCodingVisionFunction callingTool useAgentic

1473.6

Elo · rank #127

Parameters

200B

Active params

20B (MoE)

Context

256K tokens

Architecture

Sparse MoE, 200B total / 20B active (7 experts/token, ~140B-dense-equivalent)

License

Proprietary

Languages

—

API price (in/out)

$0.11 / $0.275

Modalities

text

Benchmark results

Bar shows position within the tracked field; marker = field best

BIG-Bench HardReasoning91.6%#4

best: ERNIE 4.5 300B-A47B · 94.3%

C-EvalKnowledge91.8%#9

best: Qwen3.6-Plus · 93.3%

CMMLUKnowledge90.9%#1

best: this model · 90.9%

DROPReasoning93#2

best: Hunyuan-T1 · 93.1

GPQA DiamondReasoning65.0%#147

best: GPT-5.6 · 94.6%

IFEvalReasoning89.5%#22

best: Gemma 4 26B A4B · 98.5%

MATH-500Math88.6%#52

best: GPT-5 · 99.4%

MBPP+Coding78.0%#8

best: Llama 3.1 405B · 88.6%

MMLU-ProKnowledge80.1%#56

best: Claude Fable 5 · 91.5%

MMLUKnowledge88.6%#21

best: OpenAI o3 · 92.9%

OlympiadBenchMath59.8%#1

best: this model · 59.8%

SuperGPQAReasoning55.1%#18

best: Qwen3.7-Max · 73.6%

Run it locally

Closed weights — available via API only. No local deployment.

Input / M tok

$0.11

Output / M tok

$0.275

Doubao family

Elo progression across releases

Doubao (1st-gen)May 20241955.4 Doubao-1.5-ProJan 20251473.6 Doubao-Seed-1.6Jun 20251705.5

Resources

DOCSseed.bytedance.com APIAvailable via ByteDance API

API price $0.11/$0.275 · each benchmark row carries its own source badge (see methodology)