Google has achieved a massive 566% annual increase in processing power, now handling 3.2 quadrillion tokens per month. This milestone, revealed at Google I/O, coincides with the debut of Gemini Omni, a significant bridge toward Artificial General Intelligence (AGI). Unlike standard models, Omni seamlessly merges video, imagery, and interactive simulations with physical modeling to generate content from any prompt.
The announcement also highlighted technical leaps in efficiency and agency:
- Gemini 3.5 Flash: This new model outperforms the 3.1 Pro version across most benchmarks, showing specific dominance in high-speed coding tasks.
- Gemini Spark: Operating as a personal agent on dedicated Google Cloud virtual machines, this tool runs 24/7 to execute user-directed tasks.
- MCP Integration: While starting with Chrome and Gmail, Spark will eventually bridge to third-party software using the Model Context Protocol.
