Super Data Science: ML & AI Podcast with Jon Krohn 944: Gemini 3 Pro: Google’s Back on Top
70 snips
Nov 28, 2025 Google has launched its impressive Gemini 3 Pro model, making waves across various platforms. It recently outperformed competitors like GPT-5 in a range of benchmarks, showcasing significant advances in math and creative tasks. The podcast highlights its strong performance in coding and autonomous tasks, suggesting substantial improvements in screen-based applications. Companies like Wayfair and GitHub are already piloting its capabilities for enhancing infographics and AI tools. Gemini 3 Pro symbolizes Google's strategic comeback, benefiting users and developers in the rapidly evolving AI landscape.
AI Snips
Chapters
Transcript
Episode notes
Google Regains Frontier Lead
- Gemini 3 Pro appears to have reclaimed the frontier-leader position in LLM capability.
- It topped LM Arena across text, vision, coding, and web development tracks.
Big Wins On Tough Benchmarks
- Gemini 3 Pro shows large gains on hard benchmarks like Humanity's last exam and Math Arena Apex.
- It improved math performance from ~1% to ~23% versus prior rivals.
New Watermarks Across Domains
- Gemini 3 Pro set new high marks across specialized benchmarks like AIM math and scientific knowledge tests.
- It also outperformed rivals on visual reasoning and LiveCode Bench Pro.