← sources

Need a suggestion of a model/harness to extract information from plain text

source: reddit (reddit:ollama)published 2026-10-13 17:01ingested 2026-10-14 02:42

retained text

I have a multi-gigabyte mbox file and some text conversations. I would normally do this through Claude but they contain a lot of personal information. I'm trying to extract some information from all this plain text that isn't easily greppable to create an Excel report.

I have tried a few models (like qwen2.5:14b-instruct-q5_K_M, mistral-small, etc) and harnesses (open interpreter, goose) and they all start talking nonsense, refusing to run tools, writing very simple extraction scripts and then not running them, etc.

It's not like it needs a huge context, I think, since it can very well read all the emails one by one or in batches. 

I have two hardware options to run this:

* Macbook M4 Pro 24 GB (what I've been using, with ollama)

* RTX 4090 24 GB (Windows)

I need ideas because I'm at my wits' end. I have no experience with these things.

Thank you.

document fetched 2026-10-14 00:41 · canonical https://reddit.com/r/ollama/comments/1x527u5/need_a_suggestion_of_a_modelharness_to_extract