Company’s Latest AI Breakthrough Aims to Accelerate Discovery and Enhance Research Efforts

Google has announced the rollout of its most advanced AI reasoning model, Gemini 2.5 Deep Think, which is capable of answering complex questions by exploring multiple ideas simultaneously. The model, which was first unveiled at Google I/O 2025, is the company’s first publicly available multi-agent model and is set to be released to subscribers of its Ultra subscription service starting this Friday.

According to Google, Gemini 2.5 Deep Think is a significant improvement over its earlier AI models, with the company claiming that it has developed novel reinforcement learning techniques to encourage the model to make better use of its reasoning paths. The model’s capabilities include producing much longer responses than traditional AI models and automatically working with tools such as code execution and Google Search.

Google’s Gemini 2.5 Deep Think has already made a notable impact, scoring a gold medal at this year’s International Math Olympiad (IMO) using a variation of the model. The company has also released the model it used at the IMO to a select group of mathematicians and academics, with the aim of enhancing research efforts and gathering feedback on how to improve the multi-agent system for academic use cases.

Google’s Gemini 2.5 Deep Think has also demonstrated its capabilities on various benchmarks, achieving state-of-the-art performance on Humanity’s Last Exam (HLE) and outperforming AI models from OpenAI, xAI, and Anthropic on LiveCodeBench6, a challenging test of competitive coding tasks.

The company’s decision to release Gemini 2.5 Deep Think to its Ultra subscribers comes as part of its efforts to make the model accessible to a wider range of users. Google plans to share Gemini 2.5 Deep Think with a select group of testers via the Gemini API in the coming weeks, with the aim of better understanding how developers and enterprises may use its multi-agent system.

In related news, a Singapore-based deep tech startup called SixSense has developed an AI-powered platform that helps semiconductor manufacturers predict and detect potential chip defects on production lines in real-time. The platform, which has raised $8.5 million in Series A funding, is already in use at major semiconductor manufacturers like GlobalFoundries and JCET, with customers reporting up to 30% faster production cycles and a 90% reduction in manual inspection work.

The development of Gemini 2.5 Deep Think and other multi-agent models comes as the AI industry continues to converge around this approach. Several leading AI labs, including xAI and OpenAI, have already released their own multi-agent systems, with the aim of tackling complex problems and accelerating discovery.

However, the high cost of serving multi-agent systems may limit their adoption, with tech companies choosing to gate them behind their most expensive subscription plans. Google’s decision to make Gemini 2.5 Deep Think available to its Ultra subscribers may help to mitigate this issue, making the model more accessible to a wider range of users.

Overall, the rollout of Gemini 2.5 Deep Think marks a significant milestone in the development of AI and its potential applications in various fields. As the AI industry continues to evolve, it will be interesting to see how Google’s latest breakthrough will shape the future of research and discovery.