Google unveils Gemini, its largest and 'most capable' AI model

Google has launched Gemini, its most powerful AI model to date, featuring three sizes: Ultra, Pro, and Nano.

Specifically, the large language model includes three categories: the Gemini Ultra, the largest and most functional; the Gemini Pro, versatile for a variety of tasks; and the Gemini Nano, for specific tasks and mobile devices.

The model will be licensed through Google Cloud, with Gemini Pro available from December 13 via the Gemini API in Google AI Studio or Google Cloud Vertex AI.

Google showcased Gemini's ability to analyze complex data, and the Ultra model reportedly surpasses GPT-4 in benchmarks. It also outperforms human experts in MMLU (massive multitask language understanding), demonstrating nuanced understanding and reasoning across diverse subjects such as math, physics, history, law, medicine, and ethics.

While details on monetization remain unclear, Google emphasizes user experience with Gemini, and the company plans to release a technical white paper for more insights.

Gemini's launch, delayed for extensive testing, coincided with the introduction of the next-gen tensor processing unit, TPU v5p chip.

“This new era of models represents one of the biggest science and engineering efforts we’ve undertaken as a company,” CEO Sundar Pichai wrote in Wednesday’s blog post. “I’m genuinely excited for what’s ahead, and for the opportunities Gemini will unlock for people everywhere.”