Doubao Seed 2.1 Pro
ByteDance · Undisclosed (MoE)
Seven months on from 2.0: documents and charts filled in
Released 20 September 2026 as the seven-month iteration over Seed 2.0 Pro: documents rise from 96 to 97, charts from 94 to 95, video holds at 95 — composite 95.3, level with Kimi K3 on points and listed at #4. This is not a generational jump but a fix for the two weakest spots of the 2.0 cycle: long tables, cross-page comparison and multi-figure extraction. Generation slips from 88 to 87, so creative work should still look elsewhere.
Last updated Oct 18, 2026Editorial scoring
Bottom line
Eight-dimension breakdown
editorial mapping · 0–100- Doubao Seed 2.1 Pro95.3
Weights: perception 20 / documents 18 / charts 14 / video 14 / agents 14 / generation 6 / Chinese 6 / ecosystem 8. Definitions and sources on the methodology page.
Public benchmarks
as reported, not converted| Benchmark / metric | Value | Note |
|---|---|---|
| SuperCLUE-VLM 2026-09 | 91.2 | Vendor figure, not yet independently reproduced |
| OmniDocBench v1.5 | 0.098 | Edit distance, lower is better |
| 上下文 | 256K | Doubled from 2.0 Pro’s 128K |
Public scores come from different versions and prompts and are not interchangeable with our composite; they are one evidence input.
Scenario mapping
five stress tests, one scale| Scenario | Score | Observation |
|---|---|---|
| Chinese receipt OCR | 96 · Stable on long tables and handwritten forms | |
| Chart data extraction | 94 · Cross-page comparison is the main gain | |
| 30-min video QA | 95 · Still leads on temporal localization | |
| Mobile GUI control | 94 · Accurate taps; occasional backtracking on long tasks | |
| Chinese cultural context | 96 · Strong on memes and regional dishes |
Strengths and limitations
Strengths
- Documents at 97, joint first on the board; long tables and cross-page comparison are the main gains.
- Video 95 and Chinese context 96 stay top tier — the default for Chinese short video and e-commerce.
- Mature API, SDK and Volcano Engine ecosystem with grade-A evidence; the lowest integration cost.
Limitations
- Generation slips from 88 to 87; do not rely on it for creative design or image editing.
- At ¥13/1M output tokens it costs ¥2 more than 2.0 Pro; high-frequency image analysis gets expensive.
- Closed weights, no self-hosting; regulated data needs a dedicated deployment deal.
Price and real cost
List price
¥13/M tokens
≈ ¥13 / 1M output tokens. For 1,000 images at 1,024px, ~1,100 image tokens each and 300 output tokens, the monthly cost is about ¥7.9 (assuming input at 25% of output price).
Use your volumeAbout Doubao Seed 2.1 Pro
Is Doubao Seed 2.1 Pro open source?
Licence: Closed commercial API. No public weights; access is via vendor API or an on-prem licence.
How much does Doubao Seed 2.1 Pro cost?
≈ ¥13 / 1M output tokens. It ranks #4 with a composite of 95.3. Real bills depend on resolution, tiling, caching and volume discounts.
What is Doubao Seed 2.1 Pro best for?
Chinese short-video moderation and tagging; Long-table and cross-page document batches; Mobile GUI agents and automation。
Why does Doubao Seed 2.1 Pro score 95.3?
The composite weights eight dimensions: perception 20%, documents 18%, charts 14%, video 14%, agents 14%, generation 6%, Chinese 6%, ecosystem 8%. Its strongest dimension is Document parsing & Chinese OCR; its weakest is Generation & editing.
Personal ranking · Not an official score
Every ranking, score and comparison on this page is an independent personal ranking, not an official score, and represents no vendor, institution or regulator. Scores are a subjective mapping of public evidence under our own weights — change the weights and the order changes. Disagree, or spotted a factual error? Contact the site owner: jizhi2024@foxmail.com — we check each report and log corrections in the changelog.