GLM-OCR

by zai-org

#1 document OCR at 0.9B — MIT licensed, edge-deployable

520Kdl/month

0.9Bparams

HuggingFace Use in Pipeline

Identifiers

Model ID

zai-org/GLM-OCR

Feature URI

mixpeek://image_extractor@v1/zai_glm_ocr_v1

Overview

GLM-OCR is a tiny (0.9B parameter) multimodal OCR model built on the GLM-V encoder-decoder architecture. Despite its small size, it ranks #1 on OmniDocBench V1.5 (94.62 overall score), outperforming models 10x its size on complex document understanding tasks including tables, formulas, handwriting, and multi-column layouts.

Its MIT license and sub-1B parameter count make it ideal for edge deployment, serverless functions, and cost-sensitive pipelines. On Mixpeek, GLM-OCR powers document text extraction for PDFs, scanned images, and screenshots where high accuracy matters more than raw throughput.

Architecture

GLM-V encoder-decoder with vision encoder (ViT variant) and autoregressive text decoder. 0.9B total parameters. Processes document images at native resolution with adaptive tiling for multi-page documents.

Mixpeek SDK Integration

import { Mixpeek } from "mixpeek";

const mx = new Mixpeek({ apiKey: "API_KEY" });

await mx.collections.ingest({
  collection_id: "my-collection",
  source: { url: "https://example.com/document.pdf" },
  feature_extractors: [{
    name: "ocr",
    version: "v1",
    params: {
      model_id: "zai-org/GLM-OCR"
    }
  }]
});