Will Gemini’s Large Context Window Make RAG Obsolete? SullyOmar Explains

I tested the new multimodal capabilities of Google Gemini while watching SullyOmar’s YouTube video. I asked Flash: Will Gemini’s large context window make RAG (Retrieval Augmented Generation) obsolete? Here are some examples of its answers:

This YouTube video transcript, from SullyOmar channel on Feb 9, 2025, discusses the relevance of Retrieval Augmented Generation (RAG) in light of Google’s new Gemini 2.0 Flash model.

https://www.youtube.com/watch?v=iK9wefkOjkY

The main point is that traditional RAG, specifically the method of chunking and embedding single documents for question answering, might be becoming obsolete due to the advent of powerful LLMs like Gemini 2.0 Flash with massive context windows.

For multiple relevant documents, SullyOmar suggests parallel processing: filtering, sending each document separately to Gemini, and then merging answers.

Here is a code example from SullyOmar’s video how RAG can be replaced:

const { answer, reasoning, citations } = await generateObject({
  model: vertex("gemini-2.0-flash-001"),
  messages: [
    {
      role: "user",
      content: [
        {
          type: "text",
          text: "What's the net income for 2024?",
        },
        {
          type: "file",
          data: "gs://example-bucket/path/to/file with spaces.pdf",
          mimeType: "application/pdf",
        },
      ],
    },
  ],
});

Gemini Flash explains:

The video highlights the simplicity of interacting with Gemini 2.0 Flash for document-based questions with a code example. This JavaScript code snippet, likely using Google Cloud Vertex AI, demonstrates sending a question (“What’s the net income for 2024?”) along with a direct link to a PDF file in Google Cloud Storage to the `gemini-2.0-flash-001` model. The code then anticipates receiving an `answer`, `reasoning`, and `citations` from Gemini.

This example showcases how easily developers can now leverage Gemini’s large context window to directly analyze documents without needing to implement complex traditional RAG pipelines involving chunking and vector databases for single files.

Retirement Party?

The above text was the prompt suggested by Gemini 2.0 Flash for a featured image.