Source: IT Home
Google published a blog post yesterday (February 5), inviting all Gemini application users to access the latest Gemini 2.0 Flash application model and release the 2.0 Flash Thinking reasoning experimental model.
2.0 Flash: Brand new update, fully open
The 2.0 Flash model was first unveiled at the 2024 I/O conference, and quickly became a popular choice in the developer community with its low latency and high performance. The model is suitable for large-scale, high-frequency tasks, can handle context windows of up to 1 million tokens, and demonstrates strong multimodal reasoning capabilities.
The Gemini 2.0 Flash model can interact with applications including YouTube, Google Search, and Google Maps, helping users discover and expand knowledge in multiple application scenarios.
Gemini 2.0 Flash Thinking Model
The Gemini 2.0 Flash Thinking model builds on the speed and performance of 2.0 Flash, and is trained to break down prompts into a series of steps, thereby enhancing its reasoning and providing better responses.
The 2.0 Flash Thinking Experimental model shows its thinking process, allowing users to see why it responds in a certain way, what its assumptions are, and trace the model's reasoning logic. This transparency allows users to gain a deeper understanding of the model's decision-making process.

Gemini also launched a version of 2.0 Flash Thinking that interacts with apps such as YouTube, Search, and Google Maps. These connected applications have made Gemini a unique AI assistant, and in the future we will explore how to combine new reasoning capabilities with user applications to help users complete more tasks.
2.0 Pro Experimental Version: Best Programming Performance and Complex Prompt Word Processing
Google also launched the Gemini 2.0 Pro experimental version, and the official claims that the model is good at programming and can answer complex prompts. The model has a context window of 2 million tokens, which can fully analyze and understand massive amounts of information, and supports calling tools such as Google Search and code execution.
Developers can now experience this experimental model in Google AI Studio and Vertex AI, and Gemini advanced users can also access it on desktop and mobile. IT Home attaches the relevant performance comparison as follows:

2.0 Flash-Lite: The most cost-effective model
Google AI Studio also launched the Gemini 2.0 Flash-Lite model, which the official claims is the most cost-effective model to date. It aims to provide higher quality than 1.5 Flash while maintaining low cost and fast response.
The model also supports context windows of 1 million tokens and multimodal input. For example, it can generate a line of relevant descriptions for 40,000 unique photos in a paid subscription to Google AI Studio at a cost of less than $1.