News on February 6,GoogleThe company published a blog post yesterday (February 5) inviting all Gemini application users, accessing the latest Gemini 2.0 Flash application model.and liberalize the 2.0 Flash Thinking inference experiment model.
2.0 Flash: Newly updated and fully open
Originally unveiled at I/O 2024, the 2.0 Flash model has quickly become a popular choice among the developer community due to its low latency and high performance. Suitable for large-scale, high-frequency tasks, the model is capable of handling context windows of up to 1 million tokens, demonstrating powerful multimodal reasoning.
Gemini 2.0 Flash models can interact with applications such as YouTube, Google Search, and Google Maps to help users discover and expand their knowledge in a variety of application scenarios.
Gemini 2.0 Flash Thinking Models
The Gemini 2.0 Flash Thinking model builds on the speed and performance of 2.0 Flash, a model trained to break down a cue into a series of steps to enhance its reasoning and provide better responses.
2.0 Flash Thinking Experimental The model demonstrates its thinking process so that the user can see why it responds in a certain way, what its assumptions are, and trace the model's reasoning logic. This transparency allows the user to gain a deeper understanding of the model's decision-making process.
Gemini has also launched a 2.0 Flash Thinking version that interacts with apps such as YouTube, Search and Google Maps. These connected apps already make Gemini a unique AI assistant, and the future will explore how new reasoning capabilities can be combined with user apps to help users accomplish even more.
2.0 Pro Experimental: Optimal Programming Performance and Complex Prompt Word Processing
Google has also unveiled an experimental version of Gemini 2.0 Pro, a model that officials claim is good at programming and can answer complex prompts. With a context window of 2 million tokens, the model is able to comprehensively analyze and understand massive amounts of information, and supports calls to tools such as Google search and code execution.
Developers can now experience this experimental version of the model in Google AI Studio and Vertex AI, and Gemini Premium users can access it on desktop and mobile.IT Home has attached a relevant performance comparison below:
2.0 Flash-Lite: most cost-effective model
Google AI Studio has also introduced the Gemini 2.0 Flash-Lite model, which officials claim is the most cost-effective model to date. Designed to provide higher quality than 1.5 Flash while maintaining low cost and fast response.
The model also supports contextual windows and multimodal inputs for 1 million tokens, and can, for example, generate a one-line relevant description for 40,000 unique photos in a paid subscription to Google AI Studio for less than a dollar.