Google Begins Gemini 4 Pretraining Amid Rapid User Growth

Alphabet CEO Sundar Pichai announced that Google has begun its most ambitious pretraining run to date for Gemini 4, a next-generation base model designed to compete at the frontier of AI capabilities. During a Q2 2026 earnings call, Pichai outlined a new roadmap to release models at an almost monthly cadence to keep pace with rivals OpenAI and Anthropic. This pivot follows delays to the Gemini 3.5 Pro flagship reasoning model, which missed its June release date because coding performance did not meet internal goals.

Despite these setbacks, Google reports significant user growth. The Gemini app has reached 950 million monthly active users, while the AI Mode in Google Search has surpassed 1 billion monthly active users since October. Pichai credited this growth to features like Daily Brief and the Gemini Spark assistant. To improve efficiency and coding benchmarks, the company recently expanded its portfolio with Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber, though Pichai acknowledged the need for better agentic coding capabilities.

Financial performance remains strong, with nearly 90 percent of Fortune 100 companies using Gemini Enterprise. Alphabet reported a 24 percent year-on-year revenue increase, while Google Cloud revenue saw an 82 percent year-on-year surge and a record backlog exceeding $106 billion.