Gemini 2.0 is an artificial intelligence large model released by Google, an American company, on December 12, 2024.
Gemini 2.0 adopts advanced machine learning and deep learning algorithms, optimizing the neural network structure and significantly improving efficiency. In the field of natural language processing, it can deeply understand semantics and contexts, and generate natural and fluent texts, laying a solid foundation for intelligent interaction.
- Gemini 2.0 Flash: It is suitable for large-scale, high-capacity, and high-frequency tasks. It can handle a context window of up to 1 million tokens, has powerful multimodal reasoning capabilities, and supports multimodal input and text unimodal output.
- Gemini 2.0 Pro Experimental Version: It can call Google Search tools and has an extremely long context window of 2 million tokens. When writing complex codes, its data structure accuracy is extremely high. In 13 tests including general, code, and reasoning, it won 11 first places.
- Gemini 2.0 Flash – Lite: It is known as “the most cost-effective model so far”. It only costs 7.5 cents for every 1 million tokens input. It has a context window of 1 million tokens and supports multimodal input.
- Gemini 2.0 Flash Thinking Experimental Version: Developed based on the speed and performance of Gemini 2.0 Flash, it can break down instructions into a series of steps after training, strengthening the reasoning ability and providing better feedback. Currently, it only supports text and image input.
- Excellent Multimodal Capabilities: It supports native image generation, multilingual audio output, and native tool invocation. It can understand and process various inputs such as text, images, videos, and audio, and give accurate outputs.
- Rich Tool Integration: It integrates tools such as Google Search, code execution, and third-party user-defined functions, greatly expanding the application scope.
- Significant Performance Improvement: In key benchmark tests, Gemini 2.0 Flash surpasses the previous generation, Gemini 1.5 Pro, in performance, and its speed is twice as fast.
- Smart Assistants: Smart agents built based on Gemini 2.0, such as Project Astra and Jules, can achieve functions such as real-time voice and visual processing, and assist developers in solving code problems.
- Research Field: The Deep Research function provides Gemini Advanced users with powerful reasoning and long-text parsing capabilities, helping users deeply analyze complex issues and write research reports.
- Robotics Field: Gemini Robotics is an AI model specially designed for the next generation of assistive robots. It is divided into a Vision-Language-Action model and a Vision-Language model. It performs excellently in generalization and embodied reasoning benchmark tests, helping robots flexibly adapt to the environment.
- Using through Google AI Studio
- Google AI Studio is a browser-based integrated development environment provided by Google, which helps developers quickly integrate and use Google’s Gemini series of models. Domestic users can also experience many core functions of Gemini 2.0 through this platform.
- The usage steps are as follows: First, log in to the official website of Google AI Studio (aistudio.google.com) with a Google account. After successful login, enter the homepage, which covers almost all Google models. Click “create prompt” on the left panel to create a new conversation; then, in the “model” section of the right panel, select the corresponding Gemini 2.0 model according to your needs, such as “Gemini 2.0 Flash” and “Gemini 2.0 Flash – Lite”. For example, if you want to experience a model that is fast and suitable for high-capacity tasks, you can choose “Gemini 2.0 Flash”. After selecting the model, enter questions, instructions, and other contents in the dialog box, and you can interact with Gemini 2.0.
- It should be noted that Google AI Studio provides users with 50 free requests per day, but no more than 2 requests per minute, and the maximum number of tokens does not exceed 32,000. If there are too many users, the free benefits may be reduced.
- API Call
- Developers or users with certain technical skills can call Gemini 2.0 through the API. First, log in and visit the Gemini developer platform (ai.google.dev/gemini-api/…), and apply for API access permission in the console. After successfully obtaining the API Key, integrate it into your own program, website, or App. You can also use it through some supported chatboxes.
- For example, develop an intelligent customer service program and integrate the API of Gemini 2.0 into it, and use its powerful language understanding and generation capabilities to achieve efficient customer problem-solving.
- Using through Aggregation Stations like Poe
- Poe is a commonly used and useful aggregation station. New users have a chance to try Gemini for free once (the previous trial period may have been two months, and the policy may change). Complete the card binding and payment process on the Poe platform (since Gemini currently does not support domestic payment methods, you need to bind an overseas payment card. You can apply for an overseas payment card through platforms like wildcard). There is no charge for the first month, and you can experience the advantages of Gemini’s advanced models at zero cost. Before the end of the trial period, you need to decide whether to renew according to your own usage situation, otherwise the system will automatically deduct fees. If you just want to have a simple try, be sure to cancel the subscription before it expires to avoid unnecessary fees.
- Using through the Gemini Official Website and App
- Visit the official website of Gemini (gemini.google.com), and after logging in with a Google account, you can enter the Gemini Chat page. Here, you can use Gemini 1.5 Flash, try Gemini 2.0 Flash Experimental, and Gemini Advanced (which can be tried for free for one month).
- For the Gemini App, currently, Gemini 2.0 Flash is only available to Gemini Advanced subscribers. If you are a subscriber, after downloading and logging in to the Gemini App, you can experience the functions of Gemini 2.0 Flash within the app.
Different model versions of Gemini 2.0 have different pricing strategies to meet the needs of different users. The fees are mainly calculated based on the number of tokens used. Approximately 1,000 tokens are equivalent to 750 words, but it may vary depending on the text complexity.
- Gemini 2.0 Flash
- Free Quota: The API service provides a “free tier” with a low rate limit, which can be used for testing purposes. Using Gemini 2.0 Flash on Google AI Studio is completely free, and this is the case in all available countries. The free quota includes that both input and output, as well as the context cache not exceeding 1 million tokens per hour, are free.
- Pay-as-you-go: In terms of input prices, text/images/videos cost $0.1 per 1 million tokens, and audio costs $0.7 per 1 million tokens; the output text price is $0.4 per 1 million tokens; for the context cache, text/images/videos cost $0.025 per 1 million tokens, and audio costs $0.175 per 1 million tokens; the context cache (storage) costs $1 per 1 million tokens per hour.
- Gemini 2.0 Flash – Lite
- Free Quota: It also has a “free tier” in the API service with a low rate limit for testing. Using it on Google AI Studio is free, and it applies in all available countries. The free content includes input, output, and the context cache not exceeding 1 million tokens per hour.
- Pay-as-you-go: The input price, text/images/videos/audio is uniformly $0.075 per 1 million tokens; the output text price is $0.3 per 1 million tokens; the context cache costs $0.01875 per 1 million tokens; the context cache (storage) costs $1 per 1 million tokens per hour. The first 1,500 Google Search-based queries per day are free, and for the excess, it costs $35 per 1,000 queries.
- Gemini 1.5 Flash
- Free Quota: The “free tier” of the Gemini API provides free use, but there are rate limits, such as 15 requests per minute (rpm), 1 million tokens per minute (tpm), and 1,500 requests per day (rpd), and using it on Google AI Studio is free. Both input and output tokens are free within the free quota, and the context cache not exceeding 1 million tokens per hour is free.
- Pay-as-you-go: When the request volume exceeds the free quota, it is charged differently. For prompts not exceeding 128k tokens, the input costs $0.075 per 1 million tokens, the output costs $0.3 per 1 million tokens, and the context cache costs $0.01875 per 1 million tokens; for prompts exceeding 128k tokens, the input costs $0.15 per 1 million tokens, the output costs $0.6 per 1 million tokens, and the context cache costs $0.0375 per 1 million tokens; the context cache (storage) costs $1 per 1 million tokens per hour. In addition, for grounding with Google Search, it costs $35 per 1,000 requests (up to 5,000 requests per day).
- Gemini 1.5 Flash – 8B
- Free Quota: It is free to use in the “free tier” of the Gemini API, with rate limits, and using it on Google AI Studio is free. The free items include input, output, and the context cache not exceeding 1 million tokens per hour.
- Pay-as-you-go: For prompts not exceeding 128k tokens, the input costs $0.0375 per 1 million tokens, the output costs $0.15 per 1 million tokens, and the context cache costs $0.01 per 1 million tokens; for prompts exceeding 128k tokens, the input costs $0.075 per 1 million tokens, the output costs $0.3 per 1 million tokens, and the context cache costs $0.02 per 1 million tokens; the context cache (storage) costs $0.25 per 1 million tokens per hour.
