OCR models / Tencent (Hunyuan)

HunyuanOCR 1.5

Name: HunyuanOCR 1.5
Author: Tencent (Hunyuan)

Tencent (Hunyuan) · released 2026-07-06 · data checked 2026-11-06

HunyuanOCR 1.5 is a 1.1B-parameter OCR and document-parsing model from Tencent (Hunyuan) (CN), released 2026-07-06 under the Tencent Hunyuan Community License license. It scores 94.74 on OmniDocBench v1.6 (#11 of 33). It needs about 3.2 GB of memory at 16-bit.

End-to-end OCR without a separate layout step. The 1.0 weights are archived in the same repository.

OmniDocBench v1.6

94.74

#11 of 33 models

olmOCR-Bench

—

Not reported

Parameters

1.1B

OCR specialist

Memory to run

~3.2 GB

16-bit weights plus about 1 GB

Benchmarks

OmniDocBench v1.6 Lab-reported94.74

From the HunyuanOCR 1.5 paper; the leaderboard lists version 1.0 at 89.95. Best: OXR 1.0 97.03

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

Specifications

Lab: Tencent (Hunyuan) (CN)
Type: OCR / document parsing specialist
Total parameters: 1.1B
License: Tencent Hunyuan Community License
Custom Tencent license; check its terms before commercial use.
Memory to run: ~3.2 GB at 16-bit
Released: 2026-07-06

Weights and usage on Hugging Face ↗

Closest alternatives

PaddleOCR-VL 1.5 (960M, 94.93 on OmniDocBench, Apache-2.0)
GLM-OCR (1.3B, 95.22 on OmniDocBench, MIT)
MiniCPM-V 4.7 35B-A3B (35.2B, 94.24 on OmniDocBench, Apache-2.0)
Unlimited-OCR (3.3B, 94.00 on OmniDocBench, MIT)

Sources

Something wrong here? Report it.