Google has launched Gemini 2.5 Deep Think, its most advanced AI reasoning model that explores multiple ideas simultaneously before selecting the best answer. Now available to Google AI Ultra subscribers at $249.99 per month, this new technology marks a significant advance in how AI systems approach complex problems.
Deep Think uses what Google calls “parallel thinking techniques” – essentially creating multiple reasoning paths that work on a problem at the same time. This method allows the AI to generate various ideas, compare them side-by-side, and even combine or revise thoughts before delivering a final answer.
“Deep Think pushes the frontier of thinking capabilities by leveraging parallel thinking techniques,” Google explained in its announcement. The company has developed new reinforcement learning methods that encourage the model to make effective use of extended reasoning time.
The model has demonstrated impressive benchmark results, scoring 34.8% on Humanity’s Last Exam without tools – outperforming competitors like xAI’s Grok 4 (25.4%) and OpenAI’s o3 (20.3%). It also achieved 87.6% on LiveCodeBench V6, a competitive coding test, compared to Grok 4’s 79% and OpenAI’s o3’s 72%.
This consumer version of Deep Think represents a more practical implementation of technology Google previously used to achieve gold-medal results at the 2025 International Mathematical Olympiad. While that competition-grade model takes hours to process complex math problems, the consumer version is faster for everyday use but still reaches bronze-level performance on IMO benchmarks.
Similar Posts
Deep Think particularly excels at tasks requiring creativity and strategic planning. In web development testing, Google found the model produced more visually appealing and functionally superior results than standard AI systems. It also shows promise for scientific exploration, potentially helping researchers formulate mathematical conjectures or analyze complex scientific literature.
“We’ve been impressed by Deep Think’s performance on tasks that require building something complex, piece by piece,” Google noted in its blog post.
The model is currently limited to Google’s premium subscription tier. Google AI Ultra costs $249.99 monthly (with a promotional rate of $124.99 for the first three months). Subscribers can activate Deep Think in the Gemini app by selecting the 2.5 Pro model and toggling the Deep Think option, though usage is capped at a fixed number of prompts per day.
In safety testing, Gemini 2.5 Deep Think showed improved content safety compared to regular Gemini 2.5 Pro, though Google acknowledges it has “a higher tendency to refuse benign requests.” The company is conducting additional safety evaluations as it considers expanding access.
Google is also providing the full gold-medal version of Deep Think to select mathematicians and academics for research purposes, hoping to gather feedback on how the multi-agent system might enhance academic work.
For developers interested in exploring this technology, Google plans to release Deep Think with and without tools to trusted testers via the Gemini API in the coming weeks, aiming to better understand its potential for enterprise applications.