Google's Deep Think AI Model Launches with Promising Performance Against Rivals
Google launches Deep Think AI, outperforming competitors in benchmarks and enhancing user capabilities.
Key Points
- • Deep Think launched for Gemini Ultra users, costing $250 monthly.
- • Outperforms OpenAI's o3 and xAI's Grok 4 in performance benchmarks.
- • Employs 'parallel thinking' for simultaneous idea generation and processing.
- • Achieved Bronze-level performance on the 2025 IMO benchmark.
Google has officially launched the Deep Think AI model for Gemini Ultra subscribers, priced at $250 per month. This rollout follows its announcement earlier this year at the I/O developer conference, marking a significant advancement in Google's AI capabilities. Deep Think is engineered with a 'parallel thinking' architecture, allowing the model to generate and evaluate multiple concepts simultaneously, which represents a departure from traditional sequential processing methods.
The model supports various input types, such as text, images, and audio, positioning it for diverse applications in fields like academic research and enterprise problem-solving. Preliminary performance benchmarks indicate that Deep Think has outperformed major competitors, including OpenAI's o3 and xAI's Grok 4, across various tests. This achievement underlines Google's commitment to remaining at the forefront of AI technology.
Currently, access to Deep Think is exclusive to Gemini Ultra subscribers, although specific usage limits have not been disclosed. Google is also actively working on enhancing its AI offerings, with features like Deep Search and AI-assisted phone calls in development, signaling its intent to further integrate AI across its product lines. Additionally, as user feedback is collected, there may be future considerations for introducing more accessible or economical versions of Deep Think.
Deep Think has shown significant improvements over prior iterations, achieving Bronze-level performance on the 2025 International Mathematical Olympiad (IMO) benchmark. This version is notably faster and more efficient at solving complex mathematical problems, with tests indicating a substantial reduction in processing time from hours to significantly less.
The model has been utilized in specific academic settings, with select mathematicians, including prominent figures like Michel van Garrel, testing its capabilities for mathematical conjectures, alongside basic users benefiting from its enhanced usability in everyday tasks. Its design mimics human cognitive functions, allowing for strategic planning and creative problem-solving, thereby broadening its applicability across various domains, including design and algorithmic development. Overall, the launch of Deep Think represents a formidable move by Google in the competitive AI landscape, especially in relation to its rivals like OpenAI and xAI.