# What Models to use to categorize projects based on description?

**URL:** <https://discuss.huggingface.co/t/what-models-to-use-to-categorize-projects-based-on-description/66894>\
**Category:** Models\
**Created:** [December 24, 2023, 12:34pm UTC](https://discuss.huggingface.co/t/what-models-to-use-to-categorize-projects-based-on-description/66894 "2023-12-24T12:34:01Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![AndyGrat](https://avatars.discourse-cdn.com/v4/letter/a/ecb155/32.png) [@AndyGrat](https://discuss.huggingface.co/u/AndyGrat)\
**Post date:** [December 24, 2023, 12:34pm UTC](https://discuss.huggingface.co/t/what-models-to-use-to-categorize-projects-based-on-description/66894/1 "2023-12-24T12:34:01Z")

</div>

Hi all,  
I have a database of freelance projects with a long description and a title for each (beside other variables), texts are in German language. Now, I wanna categorize these projects into categories such as “Software Development” or “Product Ownership”.  
I have a training- and test-dataframe which I categorized through OpenAI’s GPT-4 API whose results are spot-on.  
Now I am looking for a free model on Hugging-Face that will come closest to the quality of the GPT-4 API which I want to fine-tune through my dataframe.  
Which model would you recommend? Bloom? RoBERTa? BART? Others?

Many thanks for any help and best regards!

---

<div class="post-metadata">

**Author:** ![psteitz](https://avatars.discourse-cdn.com/v4/letter/p/858c86/32.png) [@psteitz](https://discuss.huggingface.co/u/psteitz)\
**Post date:** [December 26, 2023, 8:36pm UTC](https://discuss.huggingface.co/t/what-models-to-use-to-categorize-projects-based-on-description/66894/2 "2023-12-26T20:36:01Z")

</div>

I would start with distilbert-base-uncased, which is fast for both training and inference. If it gives good enough results, stop there. If not, try the others you mention.