Chandra vs MiniCPM-V 4.5
Pick Chandra for scanned, multilingual documents where you need Markdown, HTML or JSON with bounding boxes, and for PDF to Markdown in RAG pipelines. Pick MiniCPM-V 4.5 if you want Apache-2.0 weights with no revenue cap, or need a quantized model on a single GPU or edge device. The main split is layout output and licence: Chandra returns coordinates but limits free commercial use, while MiniCPM-V 4.5 has no coordinates and no limits.
Chandradatalab/chandra | MiniCPM-V 4.5openbmb/minicpm-v | |
|---|---|---|
| Type | Open source | Vision model |
| Price | 2 credits / page | 2 credits / page |
| Per 1,000 pages | $2.00 | $2.00 |
| Typical speed | 2.0 s / page | 2.5 s / page |
| Best for | Complex layouts and handwriting | OCR on edge devices |
| Inputs | PDF, images | images |
| Languages | 40+ languages | 30+ languages |
| Max pages per request | 500 | 500 |
| Tables | ✓ Yes | ✓ Yes |
| Handwriting | ✓ Yes | ✓ Yes |
| Math | ✓ Yes | No |
| Bounding boxes | ✓ Yes | No |
| Confidence | No | No |
| Forms | ✓ Yes | 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 Chandra if
- It returns Markdown, HTML and JSON plus layout chunks with bounding boxes and labels per block.
- Chandra 2 scores 85.8 on olmOCR-bench and averages 77.8% across 43 languages, with numbers published by the authors.
- You can skip running GPUs yourself by using the hosted Datalab API at $1, $3 or $7.50 per 1,000 pages, with a free tier of $20/month.
Choose MiniCPM-V 4.5 if
- Weights and code are Apache-2.0, so there is no revenue or funding cap and no paid licence tier.
- It ships int4, GGUF and AWQ builds that run on llama.cpp and vLLM, which suits on-device or edge OCR.
- At 8.7B parameters it fits local document Q&A on a single GPU, and the cost is only your compute.
Key differences
| Chandra | MiniCPM-V 4.5 | |
|---|---|---|
| Licence | Code Apache-2.0; weights modified OpenRAIL-M. Free for research, personal use and companies under $2M funding/revenue. Above that you need a commercial licence. | Apache-2.0 for weights and code. |
| How you run it | Self-host with vLLM or Hugging Face, or use the hosted Datalab API. | Self-host only, with llama.cpp, vLLM, Ollama or Transformers. No hosted option, so no hosted SLA. |
| Price | Self-hosted is free within licence terms. Hosted API is $1 (fast), $3 (balanced) or $7.50 (accurate) per 1,000 pages. | Free, you pay for compute. |
| Output formats | Markdown, HTML, JSON and bounding boxes. | Plain text, Markdown and JSON. No bounding box or layout output. |
| Published benchmarks | Chandra 2: 85.8 on olmOCR-bench, 77.8% on a 43-language average, 72.7% on a 90-language average (authors' figures). | Claims leading OCRBench results and strong OmniDocBench PDF parsing, but the model card shows scores only as images. Not published as numbers. |
| Hardware | Single NVIDIA GPU via vLLM. Benchmarks run on one H100 80GB, with presets for A100, L40S and RTX 4090/5090. | 8.7B parameters with int4, GGUF and AWQ builds. No int4 VRAM figure published. |
| Known failure | Token-loop hallucinations reported on 5-6 of about 60 Russian/Kazakh scans. Weak on Urdu (43.2%), Sanskrit (51.1%) and old scans (51.1 on olmOCR-bench). | Ollama and SGLang support were still pending official releases on the model card. |
Sources
What each one is
Chandra
Datalab's OCR model that keeps layout and handles forms, tables and handwriting.
Chandra detailsMiniCPM-V 4.5
8B vision model from OpenBMB (August 2025) built to run on devices, with strong OCR and document parsing.
MiniCPM-V 4.5 detailsTry both on your own file
Run the file once with each model and compare the output.
Use one, fall back to the other
{"model": "datalab/chandra","fallbacks": ["openbmb/minicpm-v"],"file_url": "https://example.com/document.pdf"}Questions
More comparisons
Chandra alternativesMiniCPM-V 4.5 alternativesBest OCR models in 2026