On Wednesday, Google announced the general release of Gemini 2. 0, heralded as its most advanced artificial intelligence model suite to date. Previously, in December, the company provided access to developers and trusted testers, integrating some capabilities into Google products. However, this latest release is considered a "general release" by the company. The model suite features 2. 0 Flash, designed as a "workhorse model, optimal for high-volume, high-frequency tasks at scale"; 2. 0 Pro Experimental, which centers on enhancing coding performance; and 2. 0 Flash-Lite, touted as the "most cost-efficient model yet. " In terms of pricing, Gemini Flash charges developers 10 cents per million tokens for text, image, and video inputs, whereas Flash-Lite, its more economical alternative, costs just 0. 75 cents for the same inputs. These ongoing releases align with Google’s broader strategy to heavily invest in "AI agents" amid an intensifying race for dominance in artificial intelligence among both tech giants and startups. Companies such as Meta, Amazon, Microsoft, OpenAI, and Anthropic are similarly advancing toward agentic AI — models capable of executing complex, multi-step tasks on behalf of users, rather than requiring step-by-step guidance. According to Google’s December blog entry, "Over the last year, we have been investing in developing more agentic models, enabling them to understand their surroundings, think ahead, and act on your behalf with your oversight. " They added that Gemini 2. 0 introduces "new advances in multimodality, including native image and audio outputs, as well as native tool usage, " with the aim of developing new AI agents that align with their vision for a universal assistant. Anthropic, an AI startup backed by Amazon and founded by former OpenAI researchers, is a significant competitor in the pursuit of creating AI agents. In October, the company revealed that its AI agents could operate computers similarly to humans to complete intricate tasks. This capability allows Anthropic’s technology to interpret what’s on a computer screen, select options, input text, navigate websites, and perform tasks across various software platforms, including real-time internet browsing. "It's capable of using computers much like we do, " remarked Jared Kaplan, Anthropic's chief science officer, in a CNBC interview at that time, noting it can tackle tasks involving "tens or even hundreds of steps. " Recently, OpenAI introduced a similar tool named Operator, designed to automate tasks such as organizing vacations, filling out forms, making restaurant reservations, and grocery shopping.

The Microsoft-backed company described Operator as "an agent that can access the web to perform tasks for you. " Earlier this week, OpenAI unveiled another tool called Deep Research, which allows an AI agent to compile in-depth research reports and analyze questions and topics selected by the user. Google also launched a similarly named Deep Research in December, which serves as a "research assistant to explore complex topics and compile reports on your behalf. " According to CNBC's December report, Google intends to roll out several AI features in early 2025. "In history, being first isn't always essential, but executing well and being the best in your category is crucial, " stated CEO Sundar Pichai during a strategy meeting at that time. "I believe that’s the essence of what 2025 will be about. "