GLM-OCR
- In Input modalities: ImagePDF
- Structured output
- Open weights
Prices
2 offerings · prices as listedGLM-OCR| Conditions/ 1M tokens | Input | Output |
|---|---|---|
| Standard | $0.030 | $0.030 |
Prices from Pricing (opens in a new tab) (checked Nov 1, 2026)
GLM-OCR| Conditions/ 1M tokens | Input | Output |
|---|---|---|
| Standard | ¥0.20 | ¥0.20 |
Prices from 智谱 BigModel：API 定价 (opens in a new tab) (checked Nov 2, 2026)
Standard: the price a request with no conditions pays (the headline price elsewhere on the site).
Details
- File formats
- PDF, jpg, png
- Output formats
- html
- Max file size (MB)
- 50
- Licence
- MIT License
- Model ID
Not stated on Zhipu AI’s pages: context, max output, languages.
Run locally
Weights: zai-org/GLM-OCR- Parameters 1.3B
- Dense
- Native precision BF16
- Layers 16
- KV cache per token 64 KB
- Config positions 128K (config.json max_position_embeddings)
- Needs
- ≈ 3.5 GB (weights 2.5 (maker’s file) + KV cache 0.5 + overhead 0.5)
- Usable
- 21.6 GB (90% of its memory)
- Generation speed
- 158–237 tok/s
- Longest context
- 128K (the model’s limit)
- First token
- about 0.04–0.06 s for a 2K prompt
| Precision | Weights | Needs at 8K | On your hardware | Longest context |
|---|---|---|---|---|
| BF16 (native)maker’s file | 2.5 GB | ≈ 3.5 | ●Runs well 158–237 tok/s | 128K |
| Q8_0 | 1.3 GB | ≈ 2.3 | ●Runs well 259–389 tok/s | 128K |
| FP8 | 1.2 GB | ≈ 2.2 | ●Runs well 271–406 tok/s | 128K |
| Q6_K | 1.0 GB | ≈ 2.0 | ●Runs well 310–465 tok/s | 128K |
| Q5_K_M | 0.9 GB | ≈ 1.9 | ●Runs well 341–511 tok/s | 128K |
| Q5_0 | 0.8 GB | ≈ 1.8 | ●Runs well 348–522 tok/s | 128K |
| Q4_K_M | 0.7 GB | ≈ 1.7 | ●Runs well 376–563 tok/s | 128K |
| Q4_0 | 0.7 GB | ≈ 1.7 | ●Runs well 393–590 tok/s | 128K |
| IQ4_XS | 0.7 GB | ≈ 1.7 | ●Runs well 406–609 tok/s | 128K |
| Q3_K_Mquality drops | 0.6 GB | ≈ 1.6 | ●Runs well 425–638 tok/s | 128K |
| IQ3_XXSquality drops | 0.5 GB | ≈ 1.5 | ●Runs well 483–724 tok/s | 128K |
| Q2_Kquality drops | 0.5 GB | ≈ 1.5 | ●Runs well 488–731 tok/s | 128K |
| IQ2_XXSquality drops | 0.3 GB | ≈ 1.3 | ●Runs well 574–861 tok/s | 128K |
Click a row to choose a precision. “Maker’s file” sizes are the maker’s own files; others are parameters × bits per weight.
Compared with other hardware
Each at the highest precision it runs. Click one for details.
Smallest that runs it (Q4_K_M, 8K): GPU GeForce RTX 3080; unified memory Apple M5 · 16 GB.
Run commands
vllm serve zai-org/GLM-OCR --max-model-len 8192
Memory to fine-tune
Batch 1, 2,048-token sequences, gradient checkpointing; LoRA and QLoRA train rank-16 adapters on the attention projections.
Estimates; facts from the maker’s Hugging Face repositories, as of 2026-11-06. How it’s estimated
Cited scores
From third-party leaderboards, not measured by usNo cited leaderboard covers OCR models yet. We will cite one when its license allows.
Changes
All changes →- 1 new price on Z.ai API Output $0.030
- 1 new price on Z.ai API, China Output ¥0.20
Sources
- 智谱 BigModel：API 定价in Chinese (opens in a new tab) docs.bigmodel.cn · checked Nov 2, 2026
- Pricing (opens in a new tab) docs.z.ai · checked Nov 1, 2026
- GLM-OCR - Overview - Z.AI DEVELOPER DOCUMENT (opens in a new tab) docs.z.ai · checked Nov 4, 2026
- New Released (opens in a new tab) docs.z.ai · checked Oct 5, 2026
- GLM-OCR - 智谱AI开放文档in Chinese (opens in a new tab) docs.bigmodel.cn · checked Nov 3, 2026
- zai-org/GLM-OCR model card (opens in a new tab) huggingface.co · checked Oct 31, 2026
- zai-org/GLM-OCR config.json (opens in a new tab) huggingface.co · checked Nov 1, 2026
- zai-org/GLM-OCR files (opens in a new tab) huggingface.co · checked Oct 31, 2026
- GitHub - zai-org/GLM-OCR: GLM-OCR: Accurate × Fast × Comprehensive · GitHub (opens in a new tab) github.com · checked Oct 30, 2026
Zhipu AI’s official policies: data use, privacy, terms →
Zhipu AI · China · z.ai
Related models
Snapshots, the same family and other models from this provider