Nanonets-OCR2 3B

Name: Nanonets-OCR2 3B
Author: Nanonets

Nanonets · released 2025-10-13 · data checked 2026-11-06

Nanonets-OCR2 3B is a 3.8B-parameter OCR and document-parsing model from Nanonets (US), released 2025-10-13 under the Not stated license. It scores 69.5 on olmOCR-Bench (#23 of 24). It needs about 8.5 GB of memory at 16-bit.

Built on Qwen2.5-VL 3B; strong at forms, checkboxes, signatures and watermarks.

OmniDocBench v1.6

—

Not reported

olmOCR-Bench

69.5

#23 of 24 models

Parameters

3.8B

OCR specialist

Memory to run

~8.5 GB

16-bit weights plus about 1 GB

Benchmarks

olmOCR-Bench Third party69.5

Measured by CodeSOTA. Best: OXR 1.0 87.8

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

Specifications

Lab: Nanonets (US)
Type: OCR / document parsing specialist
Total parameters: 3.8B
License: Not stated
No license is stated on the model card.
Memory to run: ~8.5 GB at 16-bit
Released: 2025-10-13

Weights and usage on Hugging Face ↗

Closest alternatives

FireRed-OCR (2.1B, 70.2 on olmOCR-Bench, Apache-2.0)
MonkeyOCR-pro 1.2B (1.2B, 71.8 on olmOCR-Bench, Non-commercial)
GLM-OCR (1.3B, 75.2 on olmOCR-Bench, MIT)
MinerU2.5 (2509) (1.2B, 75.2 on olmOCR-Bench, AGPL-3.0)

Sources

Something wrong here? Report it.