Doubao Seed 2.0 Pro

doubaodoubao/doubao-seed-2-0-pro

Doubao Seed 2.0 Pro — flagship Seed 2.0 model with strongest reasoning and tool use.

ByteDance Seed 2.0 announcement ↗Seed2.0 Model Card (PDF) ↗ByteDance Seed product page ↗Volcano Engine model pricing ↗

Doubao-Seed-2.0 Pro is ByteDance's flagship general-purpose model in the Seed 2.0 family, released February 14, 2026. It is positioned for complex reasoning, long-horizon agent workflows, and multimodal understanding. The model is served via Volcano Engine (Doubao) API and is benchmarked by ByteDance against GPT-5.2 and Gemini 3 Pro on math, coding, science, and long-video tasks.

For TheRouter operators, Seed 2.0 Pro is the strongest Chinese-origin frontier model at its price tier (~$0.47/$2.37 per million tokens via Volcano Engine). It delivers frontier-class math (AIME 2025 98.3) and competitive coding (Codeforces ~3020, SWE-Bench Verified 76.5 %), while remaining ~5–10× cheaper than GPT-5.4 / Claude Opus 4.7. The model is fully OpenAI-compatible through api.therouter.ai/v1.

Best for

• Complex reasoning, math competition problems, and long-chain agent workflows where cost-efficiency matters and Chinese-language or multimodal context is present.
• Long-video understanding, real-time video stream analysis, and time-series reasoning (VideoMME 89.5, EgoTempo above human).

Reach for something else if

• Highest-end agentic coding benchmarks (SWE-Bench Verified crown still held by Claude Opus 4.7 at 87.6 %). Route to Claude Opus 4.7 or Qwen3-Coder-Plus when maximum code agent reliability is required.
• Western enterprise procurement where ByteDance brand scrutiny (TikTok parent) is a blocker; use DeepSeek V3.2 or Qwen3-Max for lower-friction alternatives at still-competitive performance.

Context Length

131K

Max Output

33K

Input Priceper 1M tokens

$0.540/ 1M tokens

Output Priceper 1M tokens

$3.24/ 1M tokens

Modalities

text→text

Pricing Breakdown

Type	Rate
Input	$0.540 / 1M tokens
Output	$3.24 / 1M tokens

Supported Parameters

temperaturemax_tokenstop_ptoolstool_choiceresponse_formatstop

Specifications

Release date	2026-02-14technode.com ↗	verified
Developer	ByteDance Seedseed.bytedance.com ↗	verified
Training cutoff	Not publicly disclosed	unknown
License / Access	Proprietary (Volcano Engine API)www.volcengine.com ↗	verified
Context window (via TheRouter)	131072 tokens (from YAML)	verified

Benchmarks

Benchmark	Score	Source
AIME 2025 ByteDance official claim; competition-level math performance.	98.3%	tokenmix.ai ↗
Codeforces (Elo) Grandmaster tier (~top 0.1 % of competitive programmers).	3020Elo	tokenmix.ai ↗
SWE-Bench Verified Frontier-class real-world coding; behind Claude Opus 4.7 (87.6 %).	76.5%	tokenmix.ai ↗
GPQA Diamond	88.9%	tokenmix.ai ↗
VideoMME Strong long-video understanding capability.	89.5%	tokenmix.ai ↗

API Usage Examples

Use the global api.therouter.ai endpoint shown below for new integrations; the legacy China accelerated endpoint is retired.

cURL

curl https://api.therouter.ai/v1/chat/completions   -H "Content-Type: application/json"   -H "Authorization: Bearer $THE_ROUTER_API_KEY"   -d '{
    "model": "doubao/doubao-seed-2-0-pro",
    "messages": [
      {"role": "user", "content": "Summarize the key points from this input."}
    ]
  }'

API guide

Chat completion

cURL

curl https://api.therouter.ai/v1/chat/completions \
  -H "Authorization: Bearer $THEROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"doubao/doubao-seed-2-0-pro","messages":[{"role":"user","content":"Hello"}]}'

More from doubao

doubao/doubao-seed-2-0-liteCost-efficient variant of Seed 2.0 doubao/doubao-seed-2-0-miniHigh-QPS edge use cases doubao/doubao-seed-2-0-codeCoding-specialized Seed 2.0 model

Similar models

Cross-provider sibling models

deepseek/deepseek-v3.2

Strong open-weight Chinese frontier model at lower price

qwen/qwen3-235b

Alibaba's flagship reasoning model

anthropic/claude-opus-4.7

Highest SWE-Bench Verified score; premium pricing

News & changes

2026-02-14

ByteDance releases Doubao-Seed-2.0, positions Pro against GPT-5.2 and Gemini 3 Pro

ByteDance announced the Seed 2.0 family (Pro, Lite, Mini, Code) with the Pro variant claiming frontier-level math, coding, and long-video performance at ~10× lower token cost than Western flagships.

re-authored by TheRoutertechnode.com ↗

Frequently asked

How does Doubao Seed 2.0 Pro compare to DeepSeek V3.2 on reasoning and price?

Seed 2.0 Pro scores 10–20 points higher on most reasoning and coding benchmarks, but is approximately 3× more expensive than DeepSeek V3.2. Choose Seed 2.0 Pro when benchmark ceiling matters; choose DeepSeek V3.2 for maximum cost efficiency.

re-authored by TheRoutertokenmix.ai ↗

Is Doubao Seed 2.0 Pro safe to use for US/EU enterprise workloads?

ByteDance is not named in the April 2026 distillation allegations involving DeepSeek. However, ByteDance (TikTok parent) faces ongoing geopolitical scrutiny in some Western procurement processes. For consumer or non-regulated markets the model is operationally fine.

re-authored by TheRoutertokenmix.ai ↗

Fact ledger — every claim on this page traces here

source	URL	retrieved
Release date	technode.com ↗	2026-05-25	verified
Developer	seed.bytedance.com ↗	2026-05-25	verified
Training cutoff	—	—	unknown
License / Access	www.volcengine.com ↗	2026-05-25	verified
Context window (via TheRouter)	—	—	verified
AIME 2025	tokenmix.ai ↗	2026-05-25	verified
Codeforces (Elo)	tokenmix.ai ↗	2026-05-25	verified
SWE-Bench Verified	tokenmix.ai ↗	2026-05-25	verified
GPQA Diamond	tokenmix.ai ↗	2026-05-25	verified
VideoMME	tokenmix.ai ↗	2026-05-25	verified
ByteDance releases Doubao-Seed-2.0, positions Pro against GPT-5.2 and Gemini 3 Pro	technode.com ↗	2026-05-25	verified
How does Doubao Seed 2.0 Pro compare to DeepSeek V3.2 on reasoning and price?	tokenmix.ai ↗	2026-05-25	to verify
Is Doubao Seed 2.0 Pro safe to use for US/EU enterprise workloads?	tokenmix.ai ↗	2026-05-25	to verify