OCR models / Baidu (Qianfan)

Qianfan-OCR

Name: Qianfan-OCR
Author: Baidu (Qianfan)

Baidu (Qianfan) · released 2026-03-18 · data checked 2026-11-06

Qianfan-OCR is a 4.7B-parameter OCR and document-parsing model from Baidu (Qianfan) (CN), released 2026-03-18 under the Apache-2.0 license. It scores 93.90 on OmniDocBench v1.6 (#15 of 33) and 79.8 on olmOCR-Bench (#13 of 24). It needs about 10.5 GB of memory at 16-bit.

OmniDocBench v1.6

93.90

#15 of 33 models

olmOCR-Bench

79.8

#13 of 24 models

Parameters

4.7B

OCR specialist

Memory to run

~10.5 GB

16-bit weights plus about 1 GB

Benchmarks

OmniDocBench v1.6 Leaderboard93.90

Best: OXR 1.0 97.03

olmOCR-Bench Lab-reported79.8

Best: OXR 1.0 87.8

The tick marks the best score on each benchmark. See how to read the ranking.

Specifications

Lab: Baidu (Qianfan) (CN)
Type: OCR / document parsing specialist
Total parameters: 4.7B
License: Apache-2.0
Memory to run: ~10.5 GB at 16-bit
Released: 2026-03-18

Weights and usage on Hugging Face ↗

Closest alternatives

Infinity-Parser2 Pro (35.1B, 93.95 on OmniDocBench, Apache-2.0)
Unlimited-OCR (3.3B, 94.00 on OmniDocBench, MIT)
Youtu-Parsing (2.5B, 93.74 on OmniDocBench, Youtu-Parsing license)
Ovis2.6 30B-A3B (31.4B, 93.70 on OmniDocBench, Apache-2.0)

Sources

Something wrong here? Report it.