Gemini 3 Pro Preview was unveiled by Google on November 18, 2025, as part of a broader Gemini 3.0 family rollout that Google DeepMind has described as its most capable model line to date, moving beyond incremental upgrades by introducing native agentic behaviors and deeper tool integration. Rather than a traditional keynote, Google relied on a phased deployment strategy that surfaced the model first in AI Studio under the "gemini-3-pro-preview" identifier and through shadow deployments and preview endpoints, giving developers early access to its enhanced reasoning and multimodal performance. This preview release was positioned as a stepping stone for richer tool-using workflows, exemplified by its pairing with Google's Antigravity agentic developer tool on the same day.

Because Gemini 3 Pro Preview was intentionally a short-lived preview, its practical utility centers on letting teams prototype against advanced reasoning and multimodal endpoints before migrating to the production-grade successor. A published deprecation guide indicates that the preview is slated for shutdown, with developers directed to transition to Gemini 3.1 Pro to keep existing integrations intact, making the model a transitional option rather than a long-term hosting target. Teams evaluating agentic, multimodal applications can use the preview to validate tool-calling and reasoning behavior against the upcoming Gemini 3.1 baseline, while planning to retire preview-era endpoints once the migration window closes.