#49BAug 10, 2026Evidence A

PaddleOCR-VL 1.6

Name: PaddleOCR-VL 1.6 multimodal review
Item: PaddleOCR-VL 1.6
Rating: 71.5
Author: CN Multimodal Tier List

Baidu · 0.9B dense

The document-parsing ceiling at 0.9B

Baidu’s August 2026 OCR specialist: 0.9B, Apache-2.0, runnable on CPU. Documents 98 ties DeepSeek-OCR 2 for the top score here, with Chinese 92 and ecosystem 90. It is not a general multimodal model — video 40, agents 45, generation 30 — but for turning PDFs, receipts and tables into structured data, almost nothing beats it on unit cost.

TextImageBolt-on

Last updated Oct 21, 2026Editorial scoring

Bottom line

Treat it as a document-parsing component, not a multimodal model: documents 98, 0.9B and Apache-2.0 push batch OCR unit cost very low. Any video, agent or generation need requires another model; it also does not reason over complex layouts, so handwriting and poor scans still need human fallback.

Eight-dimension breakdown

editorial mapping · 0–100

PaddleOCR-VL 1.671.5

Visual perception & grounding20%

Document parsing & Chinese OCR18%

Chart & document reasoning14%

Video & long-horizon reasoning14%

GUI agents & tool use14%

Generation & editing6%

Chinese context & culture6%

Ecosystem, cost & availability8%

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
参数规模	0.9B	Official spec
许可证	Apache-2.0	Commercial use
部署	CPU	No GPU required

Sources:

PaddleOCR GitHub 飞桨官网

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		98 · Joint top tier here
Table extraction		96 · Stable on clean tables
Chart data extraction		82 · Charts are not the focus
30-min video QA		40 · No video support
Chinese cultural context		92 · Strong Chinese layouts

Strengths and limitations

Strengths

Documents 98: tied with DeepSeek-OCR 2 for the top score here.
0.9B, Apache-2.0 and CPU inference make batch OCR unit cost very low.
Chinese 92 and ecosystem 90: Chinese receipts and tables are its home turf.

Limitations

Not a general model: video 40, agents 45, generation 30.
Complex layout reasoning and handwriting still lose points.
Marked non-native: a dedicated parser, not a unified model.

Price and real cost

List price

—N/A

Open weights; runs on CPU. For 1,000 images at 1,024px, ~1,100 image tokens each and 300 output tokens, the monthly cost is about ¥0.0 (assuming input at 25% of output price).

Use your volume

About PaddleOCR-VL 1.6

Is PaddleOCR-VL 1.6 open source?

Licence: Open weights (Apache-2.0). Weights are available, but check the exact terms and filing requirements before commercial use.

How much does PaddleOCR-VL 1.6 cost?

Open weights; runs on CPU. It ranks #49 with a composite of 71.5. Real bills depend on resolution, tiling, caching and volume discounts.

What is PaddleOCR-VL 1.6 best for?

Batch PDF and receipt structuring; Table extraction; Low-cost CPU deployment。

Why does PaddleOCR-VL 1.6 score 71.5?

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.