Model

Seed1.5-Thinking

CLOSED

ByteDance · Seed family · released Apr 10, 2025

RL-trained reasoning MoE matching o3-mini-high on AIME24; served via Volcano Engine API only.

ReasoningCodingVisionFunction callingTool useAgentic

1595.3

Elo · rank #101

Parameters

200B

Active params

20B (MoE)

Context

128K tokens

Architecture

Mixture-of-Experts, 200B total / 20B active

License

Proprietary

Languages

—

API price (in/out)

$0.556 / $2.222

Modalities

text

Benchmark results

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

Aider PolyglotCoding54.2%#22

best: Claude Opus 4.5 · 89.4%

AIMEMath86.7%#61

best: GPT-5.2 · 100.0%

ARC-AGI-1Reasoning39.9%#13

best: GPT-5.6 · 96.5%

CodeforcesCoding55#46

best: DeepSeek-V4-Pro (Think Max) · 3206

GPQA DiamondReasoning77.3%#93

best: GPT-5.6 · 94.6%

IFEvalReasoning87.4%#42

best: Gemma 4 26B A4B · 98.5%

LiveCodeBenchCoding64.9%#86

best: DeepSeek-V4-Pro (Think Max) · 93.5%

MMLU-ProKnowledge87.0%#15

best: Claude Fable 5 · 91.5%

SimpleQAKnowledge12.9%#32

best: GPT-4.5 · 62.5%

SuperGPQAReasoning62.1%#14

best: Qwen3.7-Max · 73.6%

SWE-bench VerifiedCoding47.0%#95

best: Claude Fable 5 · 95.0%

Run it locally

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

Input / M tok

$0.556

Output / M tok

$2.222

Seed family

Elo progression across releases

Seed1.5-ThinkingApr 20251595.3 Seed-OSS-36BAug 20251572.9

Resources

GHByteDance-Seed/Seed-Thinking-v1.5 DOCSseed.bytedance.com/en/blog/bytedance-s-latest-thinking-model-seed-thinking-v1-5-technical-details-disclosed APIAvailable via ByteDance API

API price $0.556/$2.222 · each benchmark row carries its own source badge (see methodology)