Gemini 2.5 Pro has achieved a significant milestone by sweeping the LMArena leaderboard in all categories, showcasing its state-of-the-art performance.

The increase in tokens processed by Gemini reflects the limitless human curiosity, with a significant rise from 9.7 trillion to 480 trillion tokens per month in just a year.

The Gemini 2.5 benchmarks outperform competitors in mathematics and coding, but it is not winning the revenue race, which is a dangerous position for Google.

Gemini 2.5 Pro is our most intelligent model yet, and it excels in coding, leading on all major coding leaderboards.

Gemini 2.5 is our most secure model yet, with strengthened protections against security threats and improved transparency through thought summaries.

Gemini 1.5 is notable for being a true mixture of experts, activating specific paths for different data types, making it more energy efficient compared to models that use modular approaches.

Gemini 1.5 features a context window of one million tokens, significantly larger than GPT-4's 128,000 tokens, allowing for more extensive data processing and potentially transforming how consumer applications are built.

Gemini Diffusion is a new experimental text diffusion model that generates outputs five times faster than our previous fastest model while maintaining coding performance.

The Gemini app now has over 400 million monthly active users, with a 45% increase in usage for those using the 2.5 Pro model.