search

run-llama / llama_parse

Parse files for optimal RAG
https://www.llamaindex.ai
MIT License
2.47k stars 249 forks source link

Use of GPT 4o mini with Llama Parse #303

Closed NG-201 closed 1 month ago

NG-201 commented 2 months ago

Do you plan to use gpt 4o for Llama Parse? I think it would solve the cost issues related to Llamaparse with GPT 4O and would allow much more capabilities with graphs and complex tables compared to the "classic" LlamaParse

LLMR-boringtao commented 1 month ago

https://docs.cloud.llamaindex.ai/llamaparse/features/multimodal

NG-201 commented 1 month ago

Thanks for adding the gpt 4o mini to Llama Parse. However, I was expecting the price to be lower: It is half the GPT 4O whereas the api cost is 25x less expensive. It is not the same order of magnitude.

hexapode commented 1 month ago

@NG-201 You can provide your own API Key. However the visual token of Gpt-4o-mini are really small compare to GPT-4, so you end up with a LOT of token per page to send the same image. They advertise cheaper token but they are not the same! On a sample of 10k page we were running at half the price per page of GPT-4o with GPT-4o mini, hence the pricing. Once we have seen an higher volume of pages we may update the price.