Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
Thursday French big language model (LLM) Developer Mistral Has launched a new API for developers that manages complex PDF documents. Mistral OCR It is an optical character recognition (OCR) API that can convert a PDF to a text file so that the AI models make it easier to inhabrest.
LLMs, which include popular geni equipment like Openai Chatzipt, work especially well with raw text. So companies that want to create their own AI workflow know that it has become very important to save and index data in a clear format so that this data can be re -used for AI processing.
Unlike most OCR API, Mistral OCR is a multimodal API, which means it can detect it when there are images and photos associated with the text block. The OCR API creates a limited box around these graphical elements and includes them in the output.
The Mistral OCRO does not just give output a large wall of the text; The output is formatted in the markdown, a formatting syntax in which developers use to add links, titles and other formatting elements to a simple text file.
LLM depends a lot on the markdown for their training datasets. Similarly, when you use an AI assistant, such as Mistral Le Chat or Open AEE Chatzipt, they often create a bullet list, add links or make mardewown to make some components bold. Assistant applications format the output output in a rich text output. That’s why raw text – and MarDown – have become more important as genoy in recent years.
“Over the years, companies have often collected numerous documents in PDF or slide formats, which are accessible to LLMs, especially RAG systems. With the Mistral OCR, our customers can now convert rich and complex documents into readable materials in all languages, “Guilm Lample, co-founder and chief science officer of the misstroll.
He added, “This is an important step in accepting AI assistants in companies,” he added.
Mistral OCR is available on Mistral’s own API platform or through its cloud partners (AWS, azure, Google Cloud Vertex, etc.). And for companies working with classified or sensitive data, Mistral on-primis proposes to establish.
According to the Paris-based AI agency, the Mistral OCR performs better than the API of Google, Microsoft and Openai. The company has tested its OCR model with complex documents so that there are mathematical expressions (latex formatting), advanced layouts or tables. This is supposed to perform better with non-English documents.
Mistral OCR is just one thing and one thing, the company believes that it is faster than what is there. If you compare it to multimodal LLM like GPT -4O it is not surprising, which has OCR capacity (also contains Many Other features).
Mistral is using the Mistral OCRO for its own AI assistant CatThe When a user uploads a PDF file, the company uses the Mistral OCR in the background to understand what is in the document before processing the text.
Companies and developers will probably use the Mistral OCR with an RAG (alias recovery generation) system to use multimodal documents as input in LLM. And there are many potential use cases. For example, we can imagine the law companies using it using it to help them plow through a huge amount of documents.
RAG is a technique that is used to restore data and use it as a context with the generator AI model.
