Google Releases Advanced AI Model for Complex Reasoning Tasks
- By John K. Waters
- 08/05/25
Google has released Gemini 2.5 Deep Think, an advanced artificial intelligence model designed for complex reasoning tasks.
The model uses extended processing time to analyze multiple approaches to problems simultaneously before generating responses, Google said. Deep Think takes several minutes to produce answers compared to standard AI models that respond in seconds.
Google first previewed the technology at its I/O conference in May and has been testing it with select users. It's now available to subscribers of the company's $250-per-month AI Ultra plan. Google said the public version achieves bronze-level performance on International Mathematical Olympiad problems, while a research variant that takes hours to process queries recently earned gold-medal status.
The model scored 87.6% on LiveCodeBench, a competitive coding benchmark, compared to 80.4% in May testing. On Humanity's Last Exam, a multi-subject assessment covering advanced topics, Deep Think achieved 34.8% accuracy versus 20-25% for competing models.
AI Ultra subscribers receive limited daily queries with Deep Think through Google's Gemini interface, though the company did not specify exact usage limits. The service includes access to experimental features, increased storage, and prototype tools.
Google plans to offer Deep Think through its developer API in the coming weeks for enterprise and research applications. The model competes with OpenAI's o1 and o3 reasoning models, which use similar extended processing approaches.
The release highlights the growing competition among technology companies to develop AI systems capable of managing complex mathematical, scientific, and coding tasks. Google's premium pricing restricts access compared to standard AI models that are available for free or through lower-cost subscriptions.
Google has not disclosed adoption figures for AI Ultra or revenue from its AI services. "Because we're defining the frontier with 2.5 Pro DeepThink, we're taking extra time to conduct more frontier safety evaluations and get further input from safety experts," the DeepSeek team said in a blog post. "As part of that, we're going to make it available to trusted testers via the Gemini API to get their feedback before making it widely available."
About the Author
John K. Waters is the editor in chief of a number of Converge360.com sites, with a focus on high-end development, AI and future tech. He's been writing about cutting-edge technologies and culture of Silicon Valley for more than two decades, and he's written more than a dozen books. He also co-scripted the documentary film Silicon Valley: A 100 Year Renaissance, which aired on PBS. He can be reached at [email protected].