GLM-OCR vs InternVL3.5
Pick GLM-OCR for high-volume PDF to Markdown, academic papers and schema-based field extraction, where a 0.9B model and a $0.03 per million token API keep cost low. Pick InternVL3.5 for self-hosted document question answering, where one family from 1B to 241B-A28B lets you trade accuracy for cost. Do not rely on GLM-OCR for financial tables, since it drops footer rows and loses comma decimals.
GLM-OCRzai/glm-ocr | InternVL3.5opengvlab/internvl | |
|---|---|---|
| Type | Open source | Vision model |
| Price | 6 credits / page | 2 credits / page |
| Per 1,000 pages | $6.00 | $2.00 |
| Typical speed | 1.3 s / page | 3.0 s / page |
| Best for | Low-cost general documents | Self-hosted document understanding |
| Inputs | PDF, images | images |
| Languages | 100+ languages | Multilingual |
| Max pages per request | 500 | 500 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | No | No |
| Math | ✓ Yes | ✓ Yes |
| Bounding boxes | ✓ Yes | No |
| Confidence | ✓ Yes | No |
| Forms | No | No |
| Multilingual | ✓ Yes | ✓ Yes |
| On OCRRouter | Available | Available |
Prices are what you pay on OCRRouter. Speed is the provider's typical figure, not a benchmark we ran.
Which to choose
Choose GLM-OCR if
- It has a hosted Z.ai API at $0.03 per million tokens for input and output, or you can self-host it for free with vLLM, SGLang, Ollama or transformers.
- It scores 94.62 on OmniDocBench v1.5 per its model card, and the authors measured 1.86 PDF pages per second on a single replica.
- It can extract fields into a user-defined JSON schema, which suits forms, and it is MIT licensed.
Choose InternVL3.5 if
- It comes in nine sizes from 1B to 241B, so you can choose the accuracy and cost point that fits your hardware.
- Models up to 30B run on a single A100, and the MoE variants 20B-A4B and 30B-A3B activate few parameters per token.
- It is Apache-2.0 licensed and fits document question answering and multimodal pipelines that you host yourself.
Key differences
| GLM-OCR | InternVL3.5 | |
|---|---|---|
| Licence | MIT for the model weights; the SDK repo is Apache-2.0 | Apache-2.0 |
| How you run it | Z.ai API or self-host with vLLM, SGLang, Ollama or transformers | Self-host only, with Transformers, LMDeploy or vLLM |
| Price | API: $0.03 per million tokens for input and output. Self-hosted: free, you pay for compute | Free, self-hosted, you pay for compute |
| Model size | 0.9B | 1B to 241B-A28B; the 8B variant is 8.5B |
| OCR benchmark | OmniDocBench v1.5 94.62 and olmOCR-Bench 75.2 (both from the authors); ParseBench mean 29.6 (independent) | Not published as text; the model card shows OCR scores only as chart images |
| Layout and structure output | Markdown, JSON and plain text, plus extraction into a user-defined JSON schema | Plain text, Markdown and JSON; no bounding boxes or layout output by default |
| Known failures and limits | Repeated output without a stop sequence (issue #220), dropped table footer rows (#240), lost comma decimals (#241), no aarch64 Linux support (#227); API caps at 100 pages, 50 MB per PDF, 10 MB per image | Largest size needs eight A100 GPUs and the 38B needs two A100s; not suited to CPU-only deployment of larger sizes |
Sources
- GLM-OCR Hugging Face model card
- Z.ai GLM-OCR docs and pricing
- GLM-OCR issue #240 missing table rows
- GLM-OCR issue #241 lost decimal separators
- InternVL3.5 Hugging Face model card (8B)
- InternVL GitHub
What each one is
InternVL3.5
Shanghai AI Lab's open vision-language family (1B to 241B, August 2025), strong on OCR and document benchmarks.
InternVL3.5 detailsTry both on your own file
Run the file once with each model and compare the output.
PDF, PNG, JPG, WebP, HEIC. Up to 20 MB each, 10 files.No file?
Use one, fall back to the other
{"model": "zai/glm-ocr","fallbacks": ["opengvlab/internvl"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
GLM-OCR vs MiniCPM-V 4.5GLM-OCR vs PixtralUnlimited OCR vs InternVL3.5PaddleOCR-VL vs InternVL3.5Tesseract vs InternVL3.5OpenOCR vs InternVL3.5olmOCR vs InternVL3.5LlamaParse vs InternVL3.5
GLM-OCR alternativesInternVL3.5 alternativesBest OCR models in 2026