OCR models / Tencent (Youtu)

Youtu-Parsing

Name: Youtu-Parsing
Author: Tencent (Youtu)

Tencent (Youtu) · released 2026-01-23 · data checked 2026-11-06

Youtu-Parsing is a 2.5B-parameter OCR and document-parsing model from Tencent (Youtu) (CN), released 2026-01-23 under the Youtu-Parsing license license. It scores 93.74 on OmniDocBench v1.6 (#16 of 33). It needs about 6 GB of memory at 16-bit.

Built on Youtu-LLM 2B. Succeeded by Youtu-Parsing-Omni.

OmniDocBench v1.6

93.74

#16 of 33 models

olmOCR-Bench

—

Not reported

Parameters

2.5B

OCR specialist

Memory to run

~6 GB

16-bit weights plus about 1 GB

Benchmarks

OmniDocBench v1.6 Leaderboard93.74

Best: OXR 1.0 97.03

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

Specifications

Lab: Tencent (Youtu) (CN)
Type: OCR / document parsing specialist
Total parameters: 2.5B
License: Youtu-Parsing license
Custom Tencent license; check its terms before commercial use.
Memory to run: ~6 GB at 16-bit
Released: 2026-01-23

Weights and usage on Hugging Face ↗

Closest alternatives

Ovis2.6 30B-A3B (31.4B, 93.70 on OmniDocBench, Apache-2.0)
Qianfan-OCR (4.7B, 93.90 on OmniDocBench, Apache-2.0)
Infinity-Parser2 Pro (35.1B, 93.95 on OmniDocBench, Apache-2.0)
Unlimited-OCR (3.3B, 94.00 on OmniDocBench, MIT)

Sources

Something wrong here? Report it.