November 18, 2025 — Google launched Gemini 3 Pro Preview with stronger reasoning, multimodal understanding and agentic coding capabilities across its consumer and developer products.
What happened
Google released the first Gemini 3 model, Gemini 3 Pro Preview, on November 18, 2025. Its official developer changelog describes new controls for media resolution, thought signatures and thinking levels, while Google positioned the model for multimodal reasoning and agentic coding. The release reached the Gemini app and developer services as part of a staged rollout.
Why it matters
The launch shows how rapidly model capabilities are moving from chat interfaces into search, development and multi-step software workflows. Buyers should compare the actual task performance of available models rather than relying on a single benchmark or a vendor claim about overall intelligence.
Security and infrastructure impact
Gemini 3 also increases the importance of model governance. Google says it improved resistance to prompt injection, but customers remain responsible for access controls, data classification, monitoring and safe tool permissions in their own deployments.