Discussion about this post

User's avatar
BabbleOn's avatar

Gpt-4o is an odd choice for a RAG based system. It has a 128K context window. That limits its short term memory to ~300 pages. Gemini-1.5 has a context of 2M in comparison. Im guessing the inline context size would make a big difference in summarization.

Libby Wheeler's avatar

Yah but can it spell?

2 more comments...

Ready for more?