Google is already shifting focus toward its next flagship AI foundation model. During recent updates, CEO Sundar Pichai revealed that Gemini 4 has officially entered pre-training, describing the initiative as the largest single training run Google has ever mounted.
The announcement comes at a delicate juncture for Google’s AI strategy. Gemini 3.5 Pro, initially previewed during Google I/O, has missed its targeted summer launch and remains unavailable for general release.
Gemini 4 Pre-Training Underway
Google’s decision to begin pre-training Gemini 4 highlights an aggressive effort to reclaim momentum at the frontier level. Pichai emphasized that Google’s long-term roadmap aims for an almost monthly cadence of model updates and improvements.
According to reports covered by NokiaPowerUser, Google is devoting substantial compute power and TPU allocation toward building a significantly larger base model for Gemini 4 to compete directly with upcoming releases from OpenAI and Anthropic.
Gemini 3.5 Pro Remains Limited to Testing
Despite the excitement surrounding Gemini 4, the public rollout of Gemini 3.5 Pro remains on hold.
Announcement: First previewed during Google I/O.
Target Timeline: Originally scheduled for general availability by June.
Current Status: Restricted strictly to partner testing while Google fine-tunes reliability, software engineering capabilities, and overall stability.
Reports suggest internal benchmarks—specifically regarding complex coding performance and reasoning—fell short of expectations, prompting Google to delay the rollout rather than release an unrefined flagship model.
Flash Lineup Expands in the Interim
While work on the frontier models continues behind the scenes, Google has leaned heavily into expanding its Gemini Flash ecosystem. Recent lightweight additions (such as Gemini 3.6 Flash and Gemini 3.5 Flash-Lite) are designed to offer:
Lower latency and reduced inference overhead
Higher token efficiency for enterprise workflows
Cost-effective deployment for everyday applications
These specialized releases allow Google to demonstrate active model development while addressing the delays on its main flagship pipeline.
Looking Ahead
Google’s strategy now relies on two core promises: finalizing Gemini 3.5 Pro for broad release and laying the architectural foundation for Gemini 4. Whether Google can maintain quality and stability while pursuing a monthly release cycle will be pivotal in shaping its position in the frontier AI race.
