#226 - Gemini 3, Claude Opus 4.5, Nano Banana Pro, LeJEPA - Last Week in AI Recap

Podcast: Last Week in AI

Published: 2025-11-30

Duration: 1 hr 11 min

Summary

This episode highlights significant advancements in AI, including Google's Gemini 3 and Nano Banana Pro, showcasing impressive performance metrics and capabilities. The hosts also discuss the evolving landscape of AI tools and the competitive dynamics in the industry.

What Happened

The episode kicks off with hosts Andrei Karenkov and guest co-host Michelle Li discussing the latest updates in AI. They express enthusiasm over the recent release of Gemini 3, which achieved remarkable results on challenging benchmarks. Notably, it scored 37.4 on Humanity’s last exam, a benchmark designed to test the limits of AI capabilities. This release is seen as a pivotal moment for Google, indicating its return to the forefront of AI development, especially in light of concerns raised by OpenAI's Sam Altman regarding their competitive position.

Moving on, the conversation shifts to Nano Banana Pro, an advanced version of Google's Not a Banana model, which focuses on image editing and generation. The hosts note the impressive capabilities of Nano Banana Pro, powered by Gemini 3, particularly in its ability to perform intricate edits and solve complex problems presented in images. This release is positioned as a significant advancement in AI-driven image processing tools, showcasing the potential for more sophisticated applications in creative fields.

Key Insights

Key Questions Answered

What are the main features of Google Gemini 3?

Google Gemini 3 boasts impressive capabilities, particularly demonstrated by its record score of 37.4 on Humanity's last exam, a benchmark designed to challenge AI systems. This performance indicates that Gemini 3 is not only competitive but potentially leading in the AI space, especially following its predecessor, Gemini 2.5 Pro, which had already set high expectations earlier this year.

How does Nano Banana Pro enhance image editing?

Nano Banana Pro, the advanced version of Google’s Not a Banana model, is powered by Gemini 3, allowing for more sophisticated image editing and generation capabilities. Users can perform tasks like accurately editing elements within images while maintaining content integrity, which is a significant leap forward in AI-driven creative tools.

What concerns did Sam Altman raise about OpenAI's position?

In a leaked memo, Sam Altman expressed concerns about OpenAI potentially facing challenges due to the rapid advancements being made by Google in AI. This reflects the competitive tension in the industry as both companies strive to innovate and lead in the field of artificial intelligence, especially with the recent releases of their respective models.

What does the rollout of Gemini 3 signify for Google?

The rollout of Gemini 3 was notably smooth, reaching potentially hundreds of millions of users without any technical issues. This suggests that Google has strengthened its infrastructure capabilities, allowing it to deploy complex AI models effectively. The success of this rollout is critical as it demonstrates Google's commitment to maintaining its leadership in AI technology.

How are AI benchmarks evolving with new releases?

AI benchmarks are evolving as new models like Gemini 3 begin to outperform previous expectations. The rapid progress in AI capabilities, such as achieving higher scores on benchmarks like Humanity's last exam, indicates that LLMs are becoming increasingly sophisticated and capable of handling complex tasks that were once considered unsolvable.