# GLM-OCR by Z.ai: Open-Weight Model
Source: https://savrn.com/models/glm-ocr
Markdown alternate of the page above; the site index is https://savrn.com/llms.txt

---

## Runs On

What it takes to serve GLM-OCR (1.3B parameters): the memory its weights need at each precision, and the cheapest way to rent enough data-center GPUs to hold them.

| Precision | Weights | Memory needed | Cheapest setup | Per hour | Also fits |
| --- | --- | --- | --- | --- | --- |
| 16-bit | 2.7 GB | 3.2 GB | 1x [MI300X](https://savrn.com/ai-index/pricing/gpus/mi300x) (192 GB) Vultr | $1.85 | [1x H100](https://savrn.com/ai-index/pricing/gpus/h100) $1.99 · [1x MI325X](https://savrn.com/ai-index/pricing/gpus/mi325x) $2.00 |
| 8-bit | 1.3 GB | 1.6 GB | 1x [MI300X](https://savrn.com/ai-index/pricing/gpus/mi300x) (192 GB) Vultr | $1.85 | [1x H100](https://savrn.com/ai-index/pricing/gpus/h100) $1.99 · [1x MI325X](https://savrn.com/ai-index/pricing/gpus/mi325x) $2.00 |
| 4-bit | 0.7 GB | 0.8 GB | 1x [MI300X](https://savrn.com/ai-index/pricing/gpus/mi300x) (192 GB) Vultr | $1.85 | [1x H100](https://savrn.com/ai-index/pricing/gpus/h100) $1.99 · [1x MI325X](https://savrn.com/ai-index/pricing/gpus/mi325x) $2.00 |

Memory is the weights at that precision plus 20% for the runtime and a short context; a long context needs more. Prices are the lowest on-demand hourly rates in the [SAVRN Index](https://savrn.com/ai-index/pricing/gpus), read Oct 22, 2026.

[GLM-OCR on every accelerator the SAVRN Index prices, at every precision](https://savrn.com/models/glm-ocr/gpus)

## SAVRN's Notes on GLM-OCR

Scanned contracts, permit drawings, invoices: that is the kind of work we would hand GLM-OCR. Z.ai built it at 1.3B parameters, so the 16-bit weights take 2.7 GB and the model needs 3.2 GB to run. The cheapest setup on our board is one MI300X with 192 GB at $1.85 per hour on-demand, which leaves most of the card empty; the real question is how many copies you stack on it, not whether it fits. At 8-bit the need drops to 1.6 GB, at 4-bit to 0.8 GB.

MIT terms are short: commercial use, modification and redistribution are permitted as long as the copyright and permission notices travel with the files. Before committing, check the 131,072 token context against your longest documents, pin the September 11, 2026 revision you validate, and plan to measure per-page cost on your own hardware, since no Index host price exists yet.

SAVRN Research, 2026-09-18

## Model Card

By Z.ai, published under mit, revision 2e85a62840cc.

Join our[WeChat](https://raw.githubusercontent.com/zai-org/GLM-OCR/refs/heads/main/resources/wechat.jpg) and [Discord](https://discord.gg/QR7SARHRxK) community Use GLM-OCR's[API](https://docs.z.ai/guides/vlm/glm-ocr) [GLM-OCR SDK](https://github.com/zai-org/GLM-OCR) Recommended [Technical Report](https://arxiv.org/abs/2603.10910)

### Introduction

GLM-OCR is a multimodal OCR model for complex document understanding, built on the GLM-V encoder–decoder architecture. It introduces Multi-Token Prediction (MTP) loss and stable full-task reinforcement learning to improve training efficiency, recognition accuracy, and generalization. The model integrates the CogViT visual encoder pre-trained on large-scale image–text data, a lightweight cross-modal connector with efficient token downsampling, and a GLM-0.5B language decoder. Combined with a two-stage pipeline of layout analysis and parallel recognition based on PP-DocLayout-V3, GLM-OCR delivers robust and high-quality OCR performance across diverse document layouts.

Key Features

[Read the full model card (645 words)](https://savrn.com/models/glm-ocr/card)

## Configuration

Architecture

GlmOcrForConditionalGeneration

Context length (tokens)

131,072

Layers

16

Hidden size

1,536

Feed-forward size

4,608

Attention heads

16

Key/value heads

8

Head dimension

128

Vocabulary size

59,392

Model type

glm_ocr

## Identity and Version

Repository

zai-org/GLM-OCR

Publisher

Z.ai

Task

Image to text

Modality

Image and text

Library

transformers

Parameters

1.3B parameters

Languages

zh, en, fr, es, ru, de, ja, ko

Revision

2e85a62840ccac27daa451df36c736c4636b8628

First published

2026-01-30

Last updated

2026-09-11

## Files and Weights

12 files, 2.7 GB in total. The weights are 1 file totalling 2.7 GB in safetensors.

Weights1 file · 2.7 GB

Configuration6 files · 10.2 KB

Tokenizer2 files · 6.8 MB

Documentation1 file · 6.8 KB

Other1 file · 4.6 KB

Repository1 file · 1.5 KB

Every file

| File | Type | Size | SHA-256 |
| --- | --- | --- | --- |
| model.safetensors | Weights | 2.7 GB | a16eb0de98d1 |
| .eval_results/mdpbench.yaml | Configuration | 4.8 KB | — |
| .eval_results/olmocrbench.yaml | Configuration | 1.8 KB | — |
| config.json | Configuration | 1.7 KB | — |
| generation_config.json | Configuration | 165 B | — |
| preprocessor_config.json | Configuration | 367 B | — |
| processor_config.json | Configuration | 1.3 KB | — |
| README.md | Documentation | 6.8 KB | — |
| chat_template.jinja | Other | 4.6 KB | — |
| .gitattributes | Repository | 1.5 KB | — |
| tokenizer.json | Tokenizer | 6.8 MB | — |
| tokenizer_config.json | Tokenizer | 1.1 KB | — |

## License and Download

License

mit

Access

Open weights, no gate

Download size

2.7 GB

[Download from Z.ai](https://huggingface.co/zai-org/GLM-OCR)

Released by Z.ai through its official repository on Hugging Face. [Read the license](https://opensource.org/license/mit).

## Built From

- Described by arXiv:2603.10910

## Evaluations

Each result is shown as reported, with the conditions its reporter stated. None is a SAVRN measurement. A comparison lines two results up only when their configuration, unit and setup are all stated and identical.

| Benchmark | Conditions | Result | Reported by | Revision | Date |
| --- | --- | --- | --- | --- | --- |
| Delores-Lin/MDPBench | Task arMetric arComparison conditions not established | 21.7 | [MDPBench leaderboard](https://huggingface.co/datasets/Delores-Lin/MDPBench) Reported by a third party | Evaluated revision not stated | 2026-04-14 |
| Delores-Lin/MDPBench | Task deMetric deComparison conditions not established | 82.7 | [MDPBench leaderboard](https://huggingface.co/datasets/Delores-Lin/MDPBench) Reported by a third party | Evaluated revision not stated | 2026-04-14 |
| Delores-Lin/MDPBench | Task digitalMetric digitalComparison conditions not established | 77.9 | [MDPBench leaderboard](https://huggingface.co/datasets/Delores-Lin/MDPBench) Reported by a third party | Evaluated revision not stated | 2026-04-14 |
| Delores-Lin/MDPBench | Task enMetric enComparison conditions not established | 84.5 | [MDPBench leaderboard](https://huggingface.co/datasets/Delores-Lin/MDPBench) Reported by a third party | Evaluated revision not stated | 2026-04-14 |
| Delores-Lin/MDPBench | Task esMetric esComparison conditions not established | 75.8 | [MDPBench leaderboard](https://huggingface.co/datasets/Delores-Lin/MDPBench) Reported by a third party | Evaluated revision not stated | 2026-04-14 |
| Delores-Lin/MDPBench | Task frMetric frComparison conditions not established | 76.2 | [MDPBench leaderboard](https://huggingface.co/datasets/Delores-Lin/MDPBench) Reported by a third party | Evaluated revision not stated | 2026-04-14 |
| Delores-Lin/MDPBench | Task hiMetric hiComparison conditions not established | 39.6 | [MDPBench leaderboard](https://huggingface.co/datasets/Delores-Lin/MDPBench) Reported by a third party | Evaluated revision not stated | 2026-04-14 |
| Delores-Lin/MDPBench | Task idMetric idComparison conditions not established | 79.7 | [MDPBench leaderboard](https://huggingface.co/datasets/Delores-Lin/MDPBench) Reported by a third party | Evaluated revision not stated | 2026-04-14 |
| Delores-Lin/MDPBench | Task itMetric itComparison conditions not established | 82.8 | [MDPBench leaderboard](https://huggingface.co/datasets/Delores-Lin/MDPBench) Reported by a third party | Evaluated revision not stated | 2026-04-14 |
| Delores-Lin/MDPBench | Task jpMetric jpComparison conditions not established | 65.5 | [MDPBench leaderboard](https://huggingface.co/datasets/Delores-Lin/MDPBench) Reported by a third party | Evaluated revision not stated | 2026-04-14 |
| Delores-Lin/MDPBench | Task koMetric koComparison conditions not established | 61.2 | [MDPBench leaderboard](https://huggingface.co/datasets/Delores-Lin/MDPBench) Reported by a third party | Evaluated revision not stated | 2026-04-14 |
| Delores-Lin/MDPBench | Task latinMetric latinComparison conditions not established | 78.7 | [MDPBench leaderboard](https://huggingface.co/datasets/Delores-Lin/MDPBench) Reported by a third party | Evaluated revision not stated | 2026-04-14 |
| Delores-Lin/MDPBench | Task nlMetric nlComparison conditions not established | 80.2 | [MDPBench leaderboard](https://huggingface.co/datasets/Delores-Lin/MDPBench) Reported by a third party | Evaluated revision not stated | 2026-04-14 |
| Delores-Lin/MDPBench | Task non_latinMetric non_latinComparison conditions not established | 54.3 | [MDPBench leaderboard](https://huggingface.co/datasets/Delores-Lin/MDPBench) Reported by a third party | Evaluated revision not stated | 2026-04-14 |
| Delores-Lin/MDPBench | Task overallMetric overallComparison conditions not established | 67.3 | [MDPBench leaderboard](https://huggingface.co/datasets/Delores-Lin/MDPBench) Reported by a third party | Evaluated revision not stated | 2026-04-14 |
| Delores-Lin/MDPBench | Task photographedMetric photographedComparison conditions not established | 63.7 | [MDPBench leaderboard](https://huggingface.co/datasets/Delores-Lin/MDPBench) Reported by a third party | Evaluated revision not stated | 2026-04-14 |
| Delores-Lin/MDPBench | Task ptMetric ptComparison conditions not established | 77.4 | [MDPBench leaderboard](https://huggingface.co/datasets/Delores-Lin/MDPBench) Reported by a third party | Evaluated revision not stated | 2026-04-14 |
| Delores-Lin/MDPBench | Task ruMetric ruComparison conditions not established | 64.2 | [MDPBench leaderboard](https://huggingface.co/datasets/Delores-Lin/MDPBench) Reported by a third party | Evaluated revision not stated | 2026-04-14 |
| Delores-Lin/MDPBench | Task thMetric thComparison conditions not established | 27.4 | [MDPBench leaderboard](https://huggingface.co/datasets/Delores-Lin/MDPBench) Reported by a third party | Evaluated revision not stated | 2026-04-14 |
| Delores-Lin/MDPBench | Task viMetric viComparison conditions not established | 69.2 | [MDPBench leaderboard](https://huggingface.co/datasets/Delores-Lin/MDPBench) Reported by a third party | Evaluated revision not stated | 2026-04-14 |
| Delores-Lin/MDPBench | Task zhMetric zhComparison conditions not established | 78.5 | [MDPBench leaderboard](https://huggingface.co/datasets/Delores-Lin/MDPBench) Reported by a third party | Evaluated revision not stated | 2026-04-14 |
| Delores-Lin/MDPBench | Task zh_tMetric zh_tComparison conditions not established | 76.7 | [MDPBench leaderboard](https://huggingface.co/datasets/Delores-Lin/MDPBench) Reported by a third party | Evaluated revision not stated | 2026-04-14 |
| PaddlePaddle/Real5-OmniDocBench | Task illuminationMetric illuminationComparison conditions not established | 91.12 | [Real5-OmniDocBench Leaderboard](https://huggingface.co/datasets/PaddlePaddle/Real5-OmniDocBench) Reported by a third party | Evaluated revision not stated | 2026-08-08 |
| PaddlePaddle/Real5-OmniDocBench | Task overallMetric overallComparison conditions not established | 90.32 | [Real5-OmniDocBench Leaderboard](https://huggingface.co/datasets/PaddlePaddle/Real5-OmniDocBench) Reported by a third party | Evaluated revision not stated | 2026-08-08 |
| PaddlePaddle/Real5-OmniDocBench | Task scanningMetric scanningComparison conditions not established | 92.67 | [Real5-OmniDocBench Leaderboard](https://huggingface.co/datasets/PaddlePaddle/Real5-OmniDocBench) Reported by a third party | Evaluated revision not stated | 2026-08-08 |
| PaddlePaddle/Real5-OmniDocBench | Task screen_photographyMetric screen_photographyComparison conditions not established | 91.75 | [Real5-OmniDocBench Leaderboard](https://huggingface.co/datasets/PaddlePaddle/Real5-OmniDocBench) Reported by a third party | Evaluated revision not stated | 2026-08-08 |
| PaddlePaddle/Real5-OmniDocBench | Task skewMetric skewComparison conditions not established | 85.39 | [Real5-OmniDocBench Leaderboard](https://huggingface.co/datasets/PaddlePaddle/Real5-OmniDocBench) Reported by a third party | Evaluated revision not stated | 2026-08-08 |
| PaddlePaddle/Real5-OmniDocBench | Task warpingMetric warpingComparison conditions not established | 90.68 | [Real5-OmniDocBench Leaderboard](https://huggingface.co/datasets/PaddlePaddle/Real5-OmniDocBench) Reported by a third party | Evaluated revision not stated | 2026-08-08 |
| allenai/olmOCR-bench | Task arxiv_mathMetric arxiv_mathComparison conditions not established | 80.7 | [GLM-OCR API evaluation](https://huggingface.co/zai-org/GLM-OCR) Reported by a third party | Evaluated revision not stated | 2026-04-14 |
| allenai/olmOCR-bench | Task baselineMetric baselineComparison conditions not established | 98.8 | [GLM-OCR API evaluation](https://huggingface.co/zai-org/GLM-OCR) Reported by a third party | Evaluated revision not stated | 2026-04-14 |
| allenai/olmOCR-bench | Task headers_footersMetric headers_footersComparison conditions not established | 95.8 | [GLM-OCR API evaluation](https://huggingface.co/zai-org/GLM-OCR) Reported by a third party | Evaluated revision not stated | 2026-04-14 |
| allenai/olmOCR-bench | Task long_tiny_textMetric long_tiny_textComparison conditions not established | 86.9 | [GLM-OCR API evaluation](https://huggingface.co/zai-org/GLM-OCR) Reported by a third party | Evaluated revision not stated | 2026-04-14 |
| allenai/olmOCR-bench | Task multi_columnMetric multi_columnComparison conditions not established | 76.7 | [GLM-OCR API evaluation](https://huggingface.co/zai-org/GLM-OCR) Reported by a third party | Evaluated revision not stated | 2026-04-14 |
| allenai/olmOCR-bench | Task old_scansMetric old_scansComparison conditions not established | 37.6 | [GLM-OCR API evaluation](https://huggingface.co/zai-org/GLM-OCR) Reported by a third party | Evaluated revision not stated | 2026-04-14 |
| allenai/olmOCR-bench | Task old_scans_mathMetric old_scans_mathComparison conditions not established | 68.3 | [GLM-OCR API evaluation](https://huggingface.co/zai-org/GLM-OCR) Reported by a third party | Evaluated revision not stated | 2026-04-14 |
| allenai/olmOCR-bench | Task overallMetric overallSetup Excluding Headers & Footers category. Using ZAI API.Comparison conditions not established | 75.2 | [GLM-OCR API evaluation](https://huggingface.co/zai-org/GLM-OCR) Reported by a third party | Evaluated revision not stated | 2026-04-14 |
| allenai/olmOCR-bench | Task table_testsMetric table_testsComparison conditions not established | 77.6 | [GLM-OCR API evaluation](https://huggingface.co/zai-org/GLM-OCR) Reported by a third party | Evaluated revision not stated | 2026-04-14 |
| llamaindex/ParseBench | Task meanMetric meanSetup Pipeline name: glmocr_pipelineComparison conditions not established | 29.6 | [ParseBench](https://huggingface.co/datasets/llamaindex/ParseBench) Reported by a third party | Evaluated revision not stated | 2026-04-14 |
| llamaindex/ParseBench | Task text_contentMetric text_contentSetup Pipeline name: glmocr_pipelineComparison conditions not established | 78 | [ParseBench](https://huggingface.co/datasets/llamaindex/ParseBench) Reported by a third party | Evaluated revision not stated | 2026-04-14 |
| llamaindex/ParseBench | Task text_formattingMetric text_formattingSetup Pipeline name: glmocr_pipelineComparison conditions not established | 2.3 | [ParseBench](https://huggingface.co/datasets/llamaindex/ParseBench) Reported by a third party | Evaluated revision not stated | 2026-04-14 |

## Memory Requirements

| Precision | Weights in memory |
| --- | --- |
| As published | 2.7 GB |
| 16-bit | 2.7 GB |
| 8-bit | 1.3 GB |
| 4-bit | 0.7 GB |

Weights only, from the published parameter count; the key-value cache and runtime add to this.

## Built on This Model

- Derived from[GLM-OCR](https://savrn.com/models/glm-ocr-2)

## Questions About GLM-OCR

### How much GPU memory does GLM-OCR need?

About 3.2 GB at 16-bit and 0.8 GB at 4-bit: the weights (1.3B parameters) plus a working margin. A long context needs more.

### What is the cheapest GPU to run GLM-OCR on?

At 16-bit, 1x MI300X from $1.85 an hour; at 4-bit, 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists.

### Can I use GLM-OCR commercially?

Yes. GLM-OCR is released under MIT License. The MIT License is a short permissive license. It permits commercial use, modification and redistribution, provided the copyright notice and permission notice are included.

### What is GLM-OCR's context length?

131,072 tokens, from the maximum position embeddings in its published configuration.

## Similar Models

Model · Image to text

### [GLM-OCR](https://savrn.com/models/glm-ocr-2)

[Unsloth AI](https://savrn.com/model-publishers/unsloth)

Join our WeChat and Discord community Use GLM-OCR's API GLM-OCR is a multimodal OCR model for complex document understanding, built on the GLM-V encoder–decoder architecture. It introduces Multi-Token Prediction (MTP) loss and stable full-task reinforcement learning to improve training efficiency, recognition accuracy, and generalization. The model integrates the CogViT visual encoder pre-trained on large-scale image–text data, a lightweight cross-modal connector with efficient token downsampling, and a GLM-0.5B language decoder. Combined with a two-stage pipeline of layout analysis and parallel recognition based on PP-DocLayout-V3, GLM-OCR delivers robust and high-quality OCR performance…

Open weights mit 1.3B parameters 131,072 tokens transformers

[View model](https://savrn.com/models/glm-ocr-2)

Model · Image to text

### [GLM-OCR](https://savrn.com/models/glm-ocr-3)

[Manny](https://savrn.com/model-publishers/manny-the-one)

Join our WeChat and Discord community Use GLM-OCR's API GLM-OCR is a multimodal OCR model for complex document understanding, built on the GLM-V encoder–decoder architecture. It introduces Multi-Token Prediction (MTP) loss and stable full-task reinforcement learning to improve training efficiency, recognition accuracy, and generalization. The model integrates the CogViT visual encoder pre-trained on large-scale image–text data, a lightweight cross-modal connector with efficient token downsampling, and a GLM-0.5B language decoder. Combined with a two-stage pipeline of layout analysis and parallel recognition based on PP-DocLayout-V3, GLM-OCR delivers robust and high-quality OCR performance…

Open weights mit 1.3B parameters 131,072 tokens transformers

[View model](https://savrn.com/models/glm-ocr-3)

Model · Image to text

### [LightOnOCR-1B-1025](https://savrn.com/models/lightonocr-1b-1025)

[LightOn AI](https://savrn.com/model-publishers/lightonai)

Full BF16 version of the model. We recommend this variant for inference and further fine-tuning. LightOnOCR-1B is a compact, end-to-end vision–language model for Optical Character Recognition (OCR) and document understanding. It achieves state-of-the-art accuracy in its weight class while being several times faster and cheaper than larger general-purpose VLMs. Highlights LightOnOCR combines a Vision Transformer encoder(Pixtral-based) with a lightweight text decoder(Qwen3-based) distilled from high-quality open VLMs. It is optimized for document parsing tasks, producing accurate, layout-aware text extraction from high-resolution pages. All benchmarks evaluated using vLLM on the Olmo-Bench.…

Open weights apache-2.0 1.2B parameters 8,192 tokens transformers

[View model](https://savrn.com/models/lightonocr-1b-1025)

Model · Image to text

### [GLM-OCR-4bit](https://savrn.com/models/glm-ocr-4bit)

[MLX Community](https://savrn.com/model-publishers/mlx-community)

This model was converted to MLX format from zai-org/GLM-OCR using mlx-vlm version 0.3.10. Refer to the original model card for more details on the model.

Open weights mit 1.1B parameters 131,072 tokens transformers

[View model](https://savrn.com/models/glm-ocr-4bit)

Model · Image to text

### [kosmos-2-patch14-224](https://savrn.com/models/kosmos-2-patch14-224)

[Microsoft](https://savrn.com/model-publishers/microsoft)

This Hub repository contains a HuggingFace's transformers implementation of the original Kosmos-2 model from Microsoft. Use the code below to get started with the model. This model is capable of performing different tasks through changing the prompts. First, let's define a function to run a prompt. Here are the tasks Kosmos-2 could perform: Once you have the entities, you can use the following helper function to draw their bounding bboxes on the image

Open weights mit 1.7B parameters 2,048 tokens transformers

[View model](https://savrn.com/models/kosmos-2-patch14-224)

Model · Image to text

### [trocr-large-printed](https://savrn.com/models/trocr-large-printed)

[Microsoft](https://savrn.com/model-publishers/microsoft)

TrOCR model fine-tuned on the SROIE dataset. It was introduced in the paper TrOCR: Transformer-based Optical Character Recognition with Pre-trained Models by Li et al. and first released in this repository. Disclaimer: The team releasing TrOCR did not write a model card for this model so this model card has been written by the Hugging Face team. The TrOCR model is an encoder-decoder model, consisting of an image Transformer as encoder, and a text Transformer as decoder. The image encoder was initialized from the weights of BEiT, while the text decoder was initialized from the weights of RoBERTa. Images are presented to the model as a sequence of fixed-size patches (resolution 16x16), which…

Open weights 608M parameters transformers

[View model](https://savrn.com/models/trocr-large-printed)

## Z.ai

[All models and datasets](https://savrn.com/model-publishers/zai-org)

## Versions

- [2e85a62840cc](https://savrn.com/models/glm-ocr/versions/2e85a62840cc) · current 2026-10-22

## Explore More

- [All image to text models](https://savrn.com/models/tasks/image-to-text)
- [All models under mit](https://savrn.com/models/licenses/mit)
- [Model comparisons](https://savrn.com/models/comparisons)
- [The model directory](https://savrn.com/models)
- [Open model prices by host](https://savrn.com/ai-index/pricing/open-models)

## Source

- Repository metadata, read 2026-10-19.
- [Z.ai website](https://github.com/zai-org)
- [Hugging Face record](https://huggingface.co/zai-org/GLM-OCR)
- [How the hub is built](https://savrn.com/model-hub/methodology)