On Wednesday, Google unveiled Gemini 2.0, the latest addition to its Gemini AI model family. The new suite boasts enhanced features, including built-in support for image and audio generation. Currently, the full Gemini 2.0 model is in beta testing for a select group of developers and testers, while all users can access the Gemini 2.0 Flash AI model through web and mobile platforms. The company plans to integrate the larger model into its product offerings shortly.
Introduction of Google Gemini 2.0 AI Models
Nine months after launching the Gemini 1.5 series, Google has rolled out this upgraded version of its large language model (LLM). In a blog post, the tech giant announced the debut of the experimental Gemini 2.0 Flash model. Although this Flash version features fewer parameters, making it less suitable for intricate tasks, it makes up for these limitations with low latency and greater efficiency compared to its larger counterparts.
Google emphasized that the Gemini 2.0 Flash model now offers multimodal outputs, allowing for image generation paired with text and flexible text-to-speech (TTS) multilingual audio options. Moreover, it comes with agentic functionalities that enable the model to utilize tools such as Google Search and code execution resources, as well as third-party applications defined by users via its API.
In terms of performance, Google shared benchmark results from internal evaluations demonstrating that Gemini 2.0 Flash surpasses even the Gemini 1.5 Pro model on metrics like the Massive Multitask Language Understanding (MMLU), Natural2Code, MATH, and Graduate-Level Google-Proof Q&A (GPQA).
Users can select the experimental model from the model selector located at the top left of the web interface and at the top of the mobile app. Additionally, Gemini 2.0 is accessible via the API in Google AI Studio and Vertex AI, available to developers with support for multimodal input and text output. Currently, image generation and text-to-speech functionalities are exclusive to early-access partners of Google.
