OCR models / Alibaba (Logics)

Logics-Parsing v2

Name: Logics-Parsing v2
Author: Alibaba (Logics)

Alibaba (Logics) · released 2026-02-13 · data checked 2026-11-06

Logics-Parsing v2 is a 4.4B-parameter OCR and document-parsing model from Alibaba (Logics) (CN), released 2026-02-13 under the Not stated license. It scores 93.33 on OmniDocBench v1.6 (#18 of 33). It needs about 9.9 GB of memory at 16-bit.

A 0.8B version 3 followed in September 2026, with results only in images.

OmniDocBench v1.6

93.33

#18 of 33 models

olmOCR-Bench

—

Not reported

Parameters

4.4B

OCR specialist

Memory to run

~9.9 GB

16-bit weights plus about 1 GB

Benchmarks

OmniDocBench v1.6 Leaderboard93.33

Best: OXR 1.0 97.03

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

Specifications

Lab: Alibaba (Logics) (CN)
Type: OCR / document parsing specialist
Total parameters: 4.4B
License: Not stated
No license is stated on the model card.
Memory to run: ~9.9 GB at 16-bit
Released: 2026-02-13

Weights and usage on Hugging Face ↗

Closest alternatives

ABot-OCR (2.1B, 93.30 on OmniDocBench, Not stated)
FireRed-OCR (2.1B, 93.26 on OmniDocBench, Apache-2.0)
MinerU2.5 (2509) (1.2B, 93.04 on OmniDocBench, AGPL-3.0)
Ovis2.6 30B-A3B (31.4B, 93.70 on OmniDocBench, Apache-2.0)

Sources

Something wrong here? Report it.