
Google DeepMind has introduced Gemini 2.5 Deep Think, a cutting-edge artificial intelligence (AI) model.
Initially launched at Google I/O 2025, it is said to be the company’s first multi-agent AI model, which is particularly designed to answer questions by exploring and considering a range of ideas.
Subscribers to the company’s $250-per-month Ultra subscription can now get access to Gemini 2.5 Deep Think in the Gemini app.
The recently introduced model works differently from conventional AI-powered models by launching several AI agents in parallel on a single question.
Gemini 2.5 Deep Think uses more computational resources than single-agent models but delivers stronger results.
Notably, it has helped Alphabet-owned Google to win a gold medal at the International Math Olympiad with a variant of the model.
Google will launch that variant to select academics, noting it takes hours to reason, instead of resolving the query within a couple of minutes.
In addition, Google claimed that Gemini 2.5 Deep Think achieved better results than its competitors, including Grok AI, OpenAI and more.
Major industry players are also experimenting with multi-agent systems, including xAI (Grok 4 Heavy) and Anthropic. However, higher costs are currently limiting behind the expensive subscriptions.
It is pertinent to mention that Google plans to launch Gemini 2.5 Deep Think through the Gemini API to select developers and enterprises in the near future for broader testing.