Document AI - extractor types and pricing

[oleks_vasyliev]

Hello.

I have some questions about extractors mode and pricing. Based https://cloud.google.com/document-ai/pricing looks like "Legacy processors", like invoice/expense parsers cheaper, than "Custom extractor" in 3 times. But its called legacy. Does this mean, that this models should not be used for new extractors? Are its will be deprecated in one day, that is why its deprecated? Also, if this legacy model uptrained, is this new model still count as "Legacy processor" or it will be count as "Custom extractor" and price will be increased for it?

Thanks

[jaia · Accepted solution]

Hello,

You're right about the pricing difference between Legacy processors and Custom extractors in Document AI. Here's a breakdown of your questions:

1. Legacy processors are indeed cheaper than Custom extractors. This reflects the pre-trained nature of Legacy processors compared to the customizability of Custom extractors.
2. While "Legacy" might sound concerning, these models aren't necessarily going away soon. They are still supported and functional. However, Google Cloud might not prioritize adding new features or functionalities to Legacy processors.
3. Up-training a Legacy processor with your own data will not change its pricing category. It will remain a Legacy processor and continue to be billed under the "Legacy processor" pricing structure.

Keep an eye on Google Cloud's documentation for any updates regarding Legacy processors. While not necessarily deprecated, they might not receive the same level of development as Custom extractors.

Regards,
Jai Ade