Google DeepMind officially rolled out its highly anticipated Gemini 2.5 Deep Think model on August 1, 2025, making it available to subscribers of its Google AI Ultra plan. This advanced AI reasoning model, first unveiled at Google I/O 2025 in May, represents a significant leap in artificial intelligence capabilities, particularly in complex problem-solving and enhanced reasoning. The launch was highlighted by a tweet from Ashutosh Shrivastava, who stated, "Google DeepMind has finally launched Gemini 2.5 Deep Think."
Deep Think distinguishes itself with an experimental, enhanced reasoning mode that leverages parallel thinking techniques. This allows the model to explore multiple hypotheses and consider various solutions simultaneously before arriving at an optimal answer. This innovative approach is designed to tackle intricate challenges that require creativity, strategic planning, and step-by-step improvements, mirroring human-like deliberation.
The model has demonstrated exceptional performance across a range of demanding benchmarks. An advanced version of Gemini 2.5 Deep Think notably achieved gold-medal status at the 2025 International Mathematical Olympiad (IMO), solving complex problems end-to-end in natural language. It has also shown significant gains on competitive coding evaluations like LiveCodeBench V6 and general reasoning tests such as Humanity's Last Exam, outperforming previous Gemini iterations and some rival models.
The release of Gemini 2.5 Deep Think positions Google DeepMind firmly in the competitive artificial intelligence landscape. While the tweet from Shrivastava also alluded to future developments from competitors, stating, "Next, OpenAI will drop GPT‑5 with better benchmarks, and then you have no idea what’s coming next," Google's latest offering underscores its commitment to pushing the boundaries of AI. Access to Deep Think is currently exclusive to Google AI Ultra subscribers, priced at $250 per month.
The introduction of Deep Think is expected to provide valuable tools for developers, researchers, and academics tackling high-level computational and analytical tasks. Its ability to process complex information and generate robust solutions could accelerate breakthroughs across various scientific and technological fields. Google DeepMind plans to offer a version with and without tools to trusted testers via the Gemini API in the coming weeks.