Google rolls out Gemini 3 Deep Think to deliver advanced reasoning: How powerful is it and who can use it?

Key Highlights

• This improves its ability to handle problems that typically challenge even the strongest AI models.
• The company highlighted that Gemini 3 Deep Think performs exceptionally well on some of the toughest evaluation benchmarks.
• It reaches 41.0% on Humanity’s Last Exam without tools and an unprecedented 45.1% on ARC-AGI-2 with code execution, setting a new standard for AI-driven reasoning, adds the Google blog post.
• These advances build on the achievements of the Gemini 2.5 Deep Think variants, which recently matched gold-medal performance levels in competitions such as the International Mathematical Olympiad and the ICPC World Finals.
• Ultra subscribers can start using the new feature immediately.