---
title: "GLM-OCR | Tarka Utilities"
description: "Document vision for receipts, PDFs, screenshots, forms."
url: "https://tarkahq.com/utilities/glm-ocr/"
---

OCR

# GLM-OCR

Document vision for receipts, PDFs, screenshots, forms.

[Sign up](https://tarkahq.com/console?auth=signup) [Utility model docs](https://tarkahq.com/docs/inference/utility-models/)

- **Status**: Live
- **Languages**: General multilingual + Nepali fine-tune
- **License**: MIT base; fine-tune not declared

## Model IDs

| Model ID | Designed for |
| --- | --- |
| `glm-ocr` | General document text and structured extraction |
| `glm-ocr-nepali` | Nepali and Devanagari document text |

Confirm access with `GET /v1/models`, then follow the [utility-model docs](https://tarkahq.com/docs/inference/utility-models/) for the canonical v1 REST or native gRPC request contract.

## Pricing

`glm-ocr`: Free for a limited time. ~~$0.001~~ $0.00 per successful page

`glm-ocr-nepali`: Free for a limited time. ~~$0.001~~ $0.00 per successful page

 USD · catalog 2026-11-04.1. Release verified 2026-11-04. EUR uses the catalog conversion. [Live catalog](https://tarka.rest/control/v1/pricing).

## Overview

Tarka exposes two GLM-OCR aliases. Use `glm-ocr` for general document text and structured extraction, or `glm-ocr-nepali` for the HimalayaAI Devanagari fine-tune. Both are available through `POST /v1/ocr`, `POST /v1/chat/completions`, and native gRPC.

## What it returns

- A JSON response with the extracted `text`
- Token `usage` when reported by the model
- Server-sent events when the request sets `stream: true`

## When to pick it

- Agents that need to *read* an image — receipt → expense entry, PDF → knowledge-graph node, screenshot → structured action
- Replacing expensive per-token vision-LLM calls when the task is “extract fields”, not “reason over the image”
- Cheap first-pass OCR before falling through to a vision-LLM for ambiguous cases

Send either one base64 data URI in `image` or an ordered array in `images`. The live route does not accept remote image URLs. Use `GET /v1/models` to confirm that the alias you need is allowed for your API key.