Google Announces Gemini 2.0 with Image and Audio Support: The New 'Agentic' AI Model

In response to OpenAI's recent Sora release, Google has unveiled its latest AI model, Gemini 2.0. On Wednesday, the company introduced Gemini 2.0 Flash, the first model in the next-gen Gemini lineup. Described as a "workhorse model" for developers, Gemini 2.0 Flash offers powerful performance at scale. It supports image and audio generation, integrates seamlessly with Google Search, writes code, and is compatible with third-party apps. Alongside this launch, Google also introduced Deep Research, a feature of Gemini that browses the web to compile research reports based on user prompts.

Gemini 2.0 Flash improves upon Gemini 1.0 with enhanced reasoning, longer context windows, better understanding of complex instructions, and native tool integration. These upgrades are designed to make the model more agentic, enabling it to handle multi-step tasks on the user's behalf.

As part of this initiative, Google announced that Gemini 2.0 would be available for Project Astra, a research prototype focused on testing a universal AI assistant. Google also introduced other research prototypes: Project Mariner, aimed at exploring "human-agent interaction," and Project Jules, designed specifically for developers.

Gemini 2.0 Flash is available as an "experimental model" through the Gemini API, accessible via Google AI Studio and Vertex AI. Casual users can explore its enhanced chat features in the Gemini desktop app, with mobile app support coming soon.