Gemini 2 is Google's latest AI model, the first version of which is the fast and powerful Gemini 2 Flash. According to the company, this marks the beginning of the “agent age” in which AI can perform tasks without human input.

Google launched Gemini AI just over a year ago, ushering in a new era for the search giant that deployed AI overviews, Gemini chatbots, etc. [Gemini Flash 2 outperforms the previous version in almost all benchmarks, despite being smaller and faster.

CEO Sundar Pichai said Gemini 2 has advanced inference capabilities, and these will be included in Google Search's AI overview, providing more accurate answers to complex multi-step questions.

Gemini 2 will also likely have a Pro version with models for all Google Gemini products, including Android apps, chatbots, and experiments.

Gemini 2 has been dubbed by Google as “the age of the agent,” a model capable of advanced reasoning similar to OpenAI's o1, but also capable of natively outputting images, voice, text, etc.

The first model in the family is Gemini 2.0 Flash, but the current release is branded “experimental.” According to Google, it is twice as fast as the previous flagship model, Gemini Pro 1.5, and outperforms it in major benchmarks.

Demis Hassabis, CEO of Google DeepMind, describes Gemini 2.0 Flash as the “flagship model” with low latency and enhanced performance. Previously, Gemini was able to natively generate images, text, and audio.

It also outperforms all previous Gemini models in reasoning, is significantly better at visual comprehension, can translate speech from audio faster, and can analyze video better than Pro 1.5. Video analysis was already a special feature of Gemini.

When Gemini 1.0 was launched, we were in the era of “chatbots” that could talk to AI models and use them to create content. Then, with the advent of OpenAI o1, we entered the age of inference and at the same time the age of agents.

An agent is a term in AI where a model can create a version of itself and perform various functions on your behalf. Google also unveiled a new agent tool in Gemini that can browse the web on your behalf and return reports on complex topics.

Hassabis explains: “The practical application of AI agents is a research area full of exciting possibilities. We are exploring this new frontier with a series of prototypes that can help people accomplish tasks and get things done.”

He added that this includes Project Astra, a universal virtual assistant unveiled at Google I/O, and the new Project Mariner. It “explores the future of human-agent interaction, starting with the browser,” and includes Jules, a code agent to assist developers.

Gemini 2.0 Flash is currently available to Gemini Advanced subscribers in a drop-down model menu. It is labeled as experimental, but has been found to work without problems.

It is also available as a developer with the Gemini API or the powerful Google Gemini AI Studio. According to Google, it will be included in all Google products by 2025.