Comment by xnx

2 days ago

Given the pace AI is improving and that it doesn't give the exact same answers under many circumstances, is the the [in]stability of "preview" a concern?

GMail was in "beta" for 5 years.

4 comments

xnx

Topfi 1 day ago

Should have clarified initially what I meant by stable, especially because it isn't that known how these terms are defined for Gemini models. Not talking about getting consistent output from a not-deterministic model, but stable from a usage perspective and in the way Google uses the word "stable" to describe their model deployments [0]. "Preview" in regard to Gemini models means a few very specific restrictions including far stricter rate limits and a very tight 14 day deprecation window, making them models one cannot build on.

That is why I'd prefer for them to finish the role out of an existing model before starting work on a dedicated new version.

[0] https://ai.google.dev/gemini-api/docs/models

verdverm 2 days ago

ChatGPT 4.5 was never released to the public, but it is widely believed to be the foundation the 5.x series is built on.

Wonder how GP feels about the minor bumps for other model providers?

Topfi 1 day ago
Minor version bumps are good and I want model providers to communicate changes. The issue I am having is that Gemini "preview" class models have different deprecation timelines and rate limits, making them impossible to rely on for professional use cases. That's why I'd prefer they finish the 3.0 role out prior to putting resources into deploying a second "preview" class model.
For a stable deployment, Google needs a sufficient amount of hardware to guarantee inference and having two Pro models running makes that even more challenging: https://ai.google.dev/gemini-api/docs/models
- verdverm 1 day ago
  
  Sorry, but you come off as an armchair devops saying things like this. Google is fine, they know more than anyone else about how to run Ai at scale.
  "preview" != GA, sounds like you need to adjust your expectations