Vision AI Models (Image Input)

290 AI models on CoreAI accept images as input — you can send a photo, screenshot, chart, or scanned document and ask questions about it in plain language. Vision-capable models come from Meta, NVIDIA, ~unbiased, Unbiased, Google and other providers, and the list below is ordered newest-first from the live catalog.

Typical uses: extracting text and tables from documents, explaining diagrams, debugging code from screenshots, identifying objects, and turning whiteboard photos into structured notes. In the CoreAI app, attach an image in any chat with one of these models and it just works.

Showing the top 50 of 290 models. Browse the full directory →

Frequently Asked Questions

What is a vision AI model?

A vision (multimodal) model accepts images alongside text. Instead of describing a chart or error message in words, you attach the image and the model reads it directly — extracting text, interpreting layouts, and answering questions about what it sees.

Which AI models can read images in 2026?

290 models on CoreAI currently support image input, including the latest multimodal releases from Meta, NVIDIA, ~unbiased, Unbiased, Google. The table on this page is generated live, so it always matches what's available in the app.

Can vision AI models read handwriting and PDFs?

Most modern vision models handle printed text and clear handwriting well. For multi-page PDFs, CoreAI's Scan & PDF tool splits documents into pages a vision model can process.

Are vision models more expensive than text models?

Image input usually consumes more tokens than plain text, so a vision request costs somewhat more than a text-only one on the same model — but pricing per token is identical; you simply send more tokens.

Explore More AI Models