Google DeepMind has announced Gemini 3, positioning it as its most intelligent model to date. The company explained that this model excels in reasoning capabilities, allowing it to understand deep context and nuance.
Model ReleasesGoogle DeepMindGemini 3
Google DeepMind Releases Gemini 3 with Enhanced Reasoning Capabilities
This article is a translation. Read the Japanese original
Additionally, the company highlighted that the model can more accurately grasp user intent and return required results with fewer prompts. Google DeepMind stated that Gemini 3 delivers world-class performance in multimodal understanding.
Availability begins today. It will be provided through the Gemini app, AI Studio, Vertex AI, and Google Antigravity, a new agent development platform.
Gemini 3 has also been integrated into the AI mode of Search from day one, providing complex reasoning and a new dynamic experience.
Furthermore, Gemini 3 Deep Think, an enhanced reasoning mode, has been introduced. It will first be offered for safety testing, after which it is scheduled to be provided to Google AI Ultra subscribers.
Sources: Gemini 3 (HN 347pt, 1 comment) (HN Search (backfill), 2025-11-19)