Anthropic on Friday released another new artificial intelligence (AI) feature for its Chatbot Cloud. The feature, called PDF Image Understanding, now allows the cloud to view and process images embedded in PDF files, including charts and graphics. This capability has been added to the recently released Cloud 3.5 Sonnet AI model. The company claims that this capability will allow chatbots to accurately understand complex documents and better analyze data. Anthropic Application Packaging Interface (API) also supports PDF input. This feature is available in beta.
Anthropic releases PDF image understanding for the cloud
In its support documents, Anthropic details the new PDF support feature. The ability to understand images in PDF has been added to Cloud 3.5 Sonnet version 20241022, and it can process images to PDF as well as support PDF input.
Breaking the first capability, the cloud can now view and process images, charts and graphics added to the PDF to perform deeper analysis of the document. Once this happens, users can ask the AI questions about particular images and it can respond with relevant information.
Until now, the cloud accepted images as input and could answer queries about them, however, it could not process images attached to a document. With this feature, Anthropic now allows users to get feedback about PDFs in more detail. This feature is probably targeted towards enterprise users of the chatbot who use it to analyze sales and marketing documents as well as other such files.
Cloud 3.5 Sonnet now also accepts PDF as an input, meaning users can now directly upload PDF files and let users ask questions about them. This brings the capabilities of the cloud on par with Google’s NotebookLM, a dedicated platform for PDF and other file types.
Currently, the maximum file size of a PDF uploaded to the cloud can be 32 MB with a maximum page count of 1,000. Additionally, the chatbot cannot process PDFs that are password protected or have encryption on them. Anthropic will soon make this feature available on Amazon Bedrock and Google Vertex AI.
