Google's Gemini 3 Is Here: A Special Early Look
Hard Fork
Nov 18
Google's Gemini 3 Is Here: A Special Early Look
Google's Gemini 3 Is Here: A Special Early Look

Hard Fork
Nov 18
Google has officially unveiled Gemini 3, its most advanced large language model yet, marking a pivotal moment in the company's AI evolution. This episode dives into the technical leaps behind the new model and explores how Google plans to integrate it across its ecosystem, from search to science.
Gemini 3 sets a new benchmark in AI performance with superior reasoning, coding, and interface generation capabilities. It powers dynamic tools like interactive calculators and tutorials while being efficiently deployed at scale in Google Search. The model emphasizes practical utility, measured by real-world user assistance rather than benchmarks alone. Executives highlight ongoing integration into core products like Maps, YouTube, and Android, alongside future applications in robotics, gaming, and drug discovery. An unreleased version shows potential in transcribing historical texts with contextual understanding. Despite concerns about diminishing returns in AI scaling, the team remains confident in sustained progress, prioritizing safety—especially around function calling—and long-term value creation across industries. U.S. college students will get free premium access as part of the initial rollout.
03:12
03:12
Gemini 3 can generate interactive learning tools like a van Gogh tutorial
08:49
08:49
Gemini 3 excels in reasoning, coding, and generating intuitive interfaces
11:34
11:34
Gemini 3 is a tool to assist users in various tasks
20:06
20:06
User satisfaction matters more than benchmarks for AI products.