Confused about pricing differences between Vertex AI and Google AI Studio - especially deployment costs (Fine-tuning Gemini 2.0 Flash for low-scale text processing)

Google AI Studio
1

Hey fellow AI enthusiasts,

I’ve been diving into the world of Google’s AI offerings, and I’m a bit puzzled about the pricing differences between Vertex AI and Google AI Studio, particularly when it comes to deployment costs. I need to fine-tune Gemini 2.0 Flash for text processing on a very small scale (about 300 requests per day). Here’s what I’ve gathered so far:

  1. Google AI Studio seems cheaper for usage:
  • Input: $0.075 per million tokens
  • Output: $0.30 per million tokens
  1. Vertex AI is more expensive for usage:
  • Input: $0.15 per million tokens
  • Output: $0.60 per million tokens

But here’s where I’m confused:

  • Vertex AI has additional deployment costs, starting at $0.75 per node hour for endpoints.
  • Google AI Studio doesn’t seem to have these deployment costs.

Questions:

  1. Am I missing something about Google AI Studio’s deployment process?
  2. For those who’ve used both, how do the total costs compare in real-world usage, especially for low-volume processing?
  3. Are there hidden benefits to Vertex AI that might justify the higher costs for my small-scale use case?
  4. Any tips for minimizing deployment costs on Vertex AI given my low request volume?
  5. Can I fine-tune Gemini 2.0 Flash in Google AI Studio, or is Vertex AI my only option?

I’d really appreciate insights from those who have experience with both platforms, especially for fine-tuning and deploying models for low-volume text processing. Thanks in advance!

3

Welcome to the forum.

You can only fine-tune Gemini 1.5 flash 001 in AI Studio. The Google staff know this is a problem. I don’t know how far down in the TODO list it is to fix this.