Google has released Gemini 2.0, its new slate of artificial intelligence models and the successor to its 2024 Gemini 1.5 suite.
The models include Gemini 2.0 Flash, for high-volume tasks, 2.0 Pro Experimental, with new reasoning abilities, and 2.0 Flash-Lite, an upgraded version of 1.5 Flash that maintains its lower price point.
"Looking ahead, we’re working on more updates and improved capabilities for the Gemini 2.0 family of models,” said Koray Kavukcuoglu, CTO at Google DeepMind.
Gemini 2.0 Flash was first released in an experimental form in December, before beginning to roll out widely in the Gemini app on 30 January. Flash and Pro Experimental will now be usable in Google AI Studio and Google's Vertex AI platform, while Flash-Lite is available through these channels as a public preview.
Flash is intended to be Google’s first AI model for the “agentic era”, CEO Sundar Pichai said at the time. Agentic models are able to take action on the user’s behalf, rather than simply providing information.
According to Google, its Flash models are commonly used by developers, and are “highly capable of multimodal reasoning across vast amounts of information”. Google plans to add image generation and text-to-speech abilities to Flash.
Gemini 2.0 Pro Experimental can handle more complex prompts and has a greater understanding of general knowledge, the company said.
While 2.0 Flash-Lite will be cheaper than 2.0 Flash, it will still include 2.0 Flash’s speed and 1 million token context window.
Google also upgraded the Gemini app’s image generation capabilities in January, which the company said would add more detail to images and follow prompts more closely.
Alphabet, Google’s parent company, reported revenue of US$96.5 billion for last quarter yesterday, slightly below analysts’ expectations. Its earnings per share increased by 31% to $2.15.
Alphabet’s share price (NASDAQ: GOOG) closed at US$193.30, down from its previous close at $207.71. Its market capitalisation is $2.36 trillion.
