Gemini 3 spotted again as Google readies its release

Google is advancing preparations for Gemini 3, with direct references to Gemini 3 Pro appearing in the Enterprise tier, specifically within the agent builder’s model selector. This indicates a unified rollout strategy, where core products and agent-building tools will likely receive the update simultaneously. Such alignment would benefit enterprise teams and developers who rely on consistent access to the latest models across different workflows.

Notably, users observed that Canvas mode in Gemini’s mobile apps began producing substantially improved results compared to the web version, suggesting ongoing A/B testing or performance benchmarking. This discrepancy points toward the existence of a router model or smart deployment layer that directs queries to different model checkpoints based on platform, load, or other criteria. While some believe these changes could be due to refreshed Gemini 2.5 checkpoints, the broader context and timing suggest it is more likely tied to the upcoming Gemini 3 launch.

Current observations imply that not every prompt benefits from the new model, reinforcing the theory that only certain inputs are being routed to Gemini 3 for now, while others fall back to 2.5. For product managers, AI developers, and power users, these findings suggest that a full transition is imminent, with a broader rollout expected as soon as next week. Google’s approach highlights its ongoing effort to maintain a competitive edge by rapidly integrating its most advanced models into both consumer and enterprise Gemini offerings.