Google Gemini 3.0 - The Best LLM Yet
I’ve used Gemini 3.0 for a few hours last night and this morning for code maintenance.
Most of my effort was around software code maintenance for CEORater, some of what I asked Gemini to do was around video analytics. The adjectives that come to mind to describe Gemini 3.0 are “fast” and “accurate”. Gemini 3.0 is certainly faster and more accurate than Gemini 2.5 Pro, GPT 5.1 (OpenAI) and Claude Sonnet 4.5 (Anthropic).
Gemini 3.0 sits on top of the leaderboards this morning (HERE and HERE, two of the many leaderboards).
As I’ve written previously, Google:
has the balance sheet to constantly iterate, although 3.0 is more than a modest iteration over 2.5;
has an organic data trove as a result of Google Search, Google Photos and Gmail to name several;
and also has purpose-built, in-house chips (TPUs), that reside inside Google Cloud, both of which are optimized for Gemini. This provides Google with a sustainable advantage over OpenAI, Anthropic and xAI. Microsoft (MSFT) and Meta Platforms (META) are not in the conversation with their in-house LLMs.
If you are interested, here is a link to Google’s blog where CEO Sundar Pichai covers Gemini 3.0: Google Blog