GPT vs InternVL3.5
Pick GPT when you want a hosted API that extracts messy or handwritten documents into custom JSON and can reason over the content. Pick InternVL3.5 when you need to self-host under Apache-2.0 and choose a model size that fits your GPUs. GPT removes the infrastructure work, and InternVL3.5 gives you control over licence, hardware and cost.
GPTopenai/gpt | InternVL3.5opengvlab/internvl | |
|---|---|---|
| Type | Vision model | Vision model |
| Price | 11 credits / page | 2 credits / page |
| Per 1,000 pages | $11.00 | $2.00 |
| Typical speed | 2.4 s / page | 3.0 s / page |
| Best for | Messy scans, custom instructions | Self-hosted document understanding |
| Inputs | PDF, images | images |
| Languages | 50+ languages | Multilingual |
| Max pages per request | 100 | 500 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | ✓ Yes | No |
| Math | ✓ Yes | ✓ Yes |
| Bounding boxes | No | No |
| Confidence | No | 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 GPT if
- An HN user reported about 5% errors on messy cursive, versus about 30% for Mistral OCR, so it suits handwritten notes.
- You can prompt it to extract into custom JSON with reasoning, or to answer questions about a document.
- PDFs go in directly through the API, which sends both extracted text and page images, with files up to 50 MB per request.
Choose InternVL3.5 if
- The Apache-2.0 licence and open weights let you run it on your own hardware, and you pay only for compute.
- Nine sizes from 1B to 241B let you trade accuracy for cost, and models up to 30B fit on a single A100.
- MoE variants 20B-A4B and 30B-A3B activate few parameters per token, which suits self-hosted document question answering.
Key differences
| GPT | InternVL3.5 | |
|---|---|---|
| How you run it | Hosted OpenAI API | Self-host with Transformers, LMDeploy or vLLM |
| Licence | Proprietary | Apache-2.0, open weights |
| Pricing | Token-based per model. PDF page images add input tokens, so cost per page is not fixed. | Free software. You pay for compute. |
| Hardware | None to manage | Up to 30B on one A100, 38B on two A100s, 241B-A28B on eight A100s |
| Benchmark | GPT-5.2 scored 85.50 on OmniDocBench v1.5 and 86.59 on v1.6, versus 90.33 for Gemini 3 Pro in the same paper | No text scores. The model card publishes OCR results only as chart images. |
| Handwriting | An HN user reported about 5% errors on messy cursive | Not published |
| Coordinates | No bounding boxes or confidence scores in the standard response | No bounding boxes or layout output by default |
Sources
- OpenAI file inputs docs
- PaddleOCR-VL-1.5 paper (GPT-5.2 eval)
- HN Mistral OCR thread
- Hugging Face model card (8B)
- OpenAI API pricing
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": "openai/gpt","fallbacks": ["opengvlab/internvl"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
GPT vs MiniCPM-V 4.5GPT vs PixtralClaude vs InternVL3.5Gemini Flash vs InternVL3.5GPT vs Mathpix ConvertGPT vs TrOCRGPT vs DoclingInternVL3.5 vs Phi-4-multimodal
GPT alternativesInternVL3.5 alternativesBest OCR models in 2026