Best AI Models for Parsing Shareholder Information from PDFs and Images

[willow mesa]

@dusty flicker For safety/privacy reasons, sharing sensitive documents are prohibited

[dusty flicker]

These are public documents

You can find that under the commercial register in Germany, which is completely public

https://handelsregister.de

[willow mesa]

you could just link to the source directly, but I still recommend against reuploading it

[dusty flicker]

Ok

[dusty flicker]

Hey everyone,

I'm fairly new to the AI space and could use some support.
For a project, I want to extract information from shareholder documents (in German) and need a solid model that can analyze these documents and extract or calculate the names of the shareholders and their percentage shares. I receive the documents as PDF or PNG files.

I tried using GPT-4o-mini, but unfortunately, they don't offer fine-tuning.

Does anyone have good tips on how I could solve this problem and which model to use?

[willow mesa]

perhaps claude 3.7, grok or gemini 2.5 pro

I have tried grok to analyze a thousand lines of code

[dusty flicker]

Ahh okay, but I need to process around 10,000+ documents.
Do you have any good tips for local models or more affordable ones that still perform well? :D