MOUNTAIN VIEW, CA – Google today announced the immediate rollout of Gemini 2.5 Deep Think, a significant upgrade to its flagship artificial intelligence model, exclusively for subscribers of its new premium tier, Google AI Ultra. The announcement signals a major escalation in the competitive AI landscape, promising unprecedented reasoning and multistep task execution abilities.
Gemini 2.5 Deep Think is engineered to move beyond simple instruction-following, reportedly capable of complex logical deduction, long-term planning, and anticipating subsequent steps in a user’s workflow without explicit prompts. This “Deep Think” architecture is the result of breakthroughs in agentic AI, allowing the model to function more like an autonomous collaborator than a simple tool.
“We are not just iterating; we are redefining what’s possible in human-computer interaction,” said Demis Hassabis, CEO of Google DeepMind, in a statement. “Deep Think allows Gemini 2.5 to understand not just what you ask, but the underlying goal of your query. It can strategize, break down monumental tasks into manageable sub-tasks, and execute them with a new level of precision.”
The new model is being made available through Google AI Studio and the Vertex AI platform for developers and enterprise customers subscribed to the “Google AI Ultra” plan. This premium tier is designed for users who require the highest level of AI performance for tasks such as advanced scientific research, complex code generation, and in-depth market analysis.
According to Google, the initial rollout will be phased, with access being granted to users in North America and Europe first, followed by a wider release to the Asia-Pacific region in the coming weeks.
The launch is seen as a direct response to advancements from competitors like OpenAI and Anthropic. While benchmark performance details are still emerging, industry analysts suggest that the “Deep Think” feature is specifically designed to tackle complex, real-world problems that have previously been a challenge for generative AI models.
Google CEO Sundar Pichai added, “Our mission has always been to organize the world’s information and make it universally accessible and useful. With Gemini 2.5 Deep Think, we are taking a bold step towards making intelligence universally accessible and profoundly more useful, empowering human creativity and ingenuity like never before.”
The tech community is now eagerly awaiting independent testing to verify the new capabilities and see how Gemini 2.5 Deep Think performs in real-world applications. If its performance matches the promise, this launch could mark a pivotal moment in the evolution of artificial intelligence.
