Google Launches Its Most Capable AI Model Gemini
Gemini comes in three sizes
Tech giant Google has unveiled its new general-purpose AI model called Gemini, designed to seamlessly understand multiple data types. Unlike single model predecessors, Gemini is available in three tailored variants optimized for different uses.
Dubbed Gemini by Google, this innovative AI was trained from the outset on text, audio, images, video and more. Its three sizes each target specific applications:
Gemini Types:
Gemini Ultra: As the most powerful option, Ultra excels at highly complex enterprise tasks ideal for data centers. It surpasses existing models with state-of-the-art performance on over 30 academic benchmarks measuring skills like problem-solving. Ultra remains in testing but promises significant capabilities for intricate workloads.
Gemini Pro: Positioned as the advanced variant, Pro powers Google services and is customized for their intelligent chatbot Bard. Running on company servers, it plays a key role in advancing Bard's conversational abilities through superior reasoning, comprehension, and quick responses.
Gemini Nano: The lightweight Nano enables robust on-device AI capabilities even offline. It facilitates seamless experiences through mobile features such as summarization, transcription and suggested smart replies, ensuring responsive usage without server reliance. Nano debuted on Google Pixel 8 Pro devices.
Google CEO Sundar Pichai stated Gemini represents one of their biggest scientific undertakings. It will enhance their intelligent assistant Bard through a customized Gemini Pro version.
Consumers can experience Gemini's multimodal skills through the Pixel 8 Pro with features like summarization and smart replies. Integration continues across Search, Ads, Chrome and healthcare tool Duet AI.
Gemini achieved a 90% score surpassing experts on a 57-subject multitask test. Developers can access Gemini Pro from December via Google AI platforms. This releases its potential to develop innovative multimodal applications.
Through its distinct sizes, Gemini establishes new benchmarks for AI that seamlessly handles multiple data types, unlocking possibilities across industries.