Google on Wednesday announced the launch of Gemini, its most powerful AI model to date. The new model is available in different sizes — Nano, Pro, and Ultra — optimised for different applications, and Google hopes Gemini can compete with OpenAI’s GPT-4 model that is available for paying customers. According to the company, Gemini Nano is available on the Pixel 8 Pro and will power new on-device AI features, while Gemini Pro will be available on Google Bard, the company’s chatbot that competes with ChatGPT.
In a blog post, Alphabet CEO Sundar Pichai said that a wide-ranging effort that involved collaboration with Google Research and other teams at the company led to the development of Gemini. “It was built from the ground up to be multimodal, which means it can generalise and seamlessly understand, operate across and combine different types of information including text, code, audio, image and video,” Pichai said.
Meanwhile, both Google Bard and the company’s Search Generative Experience (SGE) service are now powered by Gemini Pro. A new version of the chatbot called Bard Advanced will use the Ultra model when it is available next year, which means that its capabilities are also expected to receive a boost. The company also said that Gemini will come to other products like Ads, Chrome, Duet AI, and Search in the coming months.
In terms of performance, it appears that Gemini is already capable of taking on the world’s leading AI models, including OpenAI’s GPT-4. A white paper released by Google shows that the top-of-the-line Gemini Ultra model beats GPT-4 in select benchmarks.
According to the company, it is also the first model that is capable of outperforming experts on the massive multitask language understanding (MMLU) dataset, which combines 57 subjects including ethics, history, law, math, medicine, and physics.
The search giant, which has lagged behind in monetising its artificial intelligence technology unlike rival OpenAI, announced access to the Gemini API in Google Studio or Google Cloud Vertex AI starting on December 13. OpenAI charges customers for access to GPT-4 and the recently unveiled GPT-4 Turbo model.
Enterprise developers and developers can pay to access Gemini Pro, via Google Cloud. Meanwhile, Gemini Nano will also be accessible to developers who want to create applications or new services on Android smartphones. There’s no word from the company on a timeline for developers and researchers to access its most powerful AI model, which is expected to arrive next year.