The Gemini 2.0 Family is now open to developers.
With the new enhancements, Gemini 2.0 is now accessible to more developers and production users. The models listed below are now accessible in Vertex AI and the Gemini API through Google AI Studio:
- With improved performance, streamlined pricing, and larger rate limits, Gemini 2.0 Flash is now widely accessible.
- The public can now preview the Gemini 2.0 Flash-Lite, Google’s newest and most affordable model to date.
- Now available is Gemini 2.0 Pro, an experimental update to its greatest model to far for sophisticated prompts and coding.
These releases enable a wide range of use cases and applications to leverage Gemini 2.0 capabilities, including its recently released Gemini 2.0 Flash Thinking Experimental, its Flash variation that thinks before responding.
Features of the model
Multimodal input, a 1 million token context window, and native tool use are just a few of the extensive feature set that Gemini 2.0 Flash provides. In addition to text output, it can also output images and audio, and in the upcoming months, the Multimodal Live API will be made publicly available. For use cases involving large-scale text production, Gemini 2.0 Flash-Lite is cost-optimized.
Gemini Model features
Performance of the model
On a variety of benchmarks, the Gemini 2.0 models significantly outperform the Gemini 1.5 versions.
Gemini 2.0 Flash, like its predecessors, defaults to a simple design that lowers costs and facilitates use. In chat-oriented use cases, it can also be encouraged to adopt a more verbose approach, which yields better results.
Gemini price
With Gemini 2.0 Flash and Gemini 2.0 Flash-Lite, it keep cutting costs. Both eliminate the Gemini 1.5 Flash distinction between short and lengthy context queries, and each input type has a single price.
This implies that, even while both 2.0 Flash and Flash-Lite offer performance gains, their costs may be less than those of Gemini 1.5 Flash for mixed-context applications.
With an industry-leading free tier and rate limits to grow to production, you can begin creating with the newest Gemini models in just four lines of code. It is impressed with your development to date and eager to learn how you’ll apply these new Gemini models. Happy construction!
Everyone can now access Gemini 2.0
Google launched an experimental version of Gemini 2.0 Flash, its incredibly effective workhorse model for developers with reduced latency and improved speed, in December, ushering in the agentic era. Its upgraded 2.0 Flash Thinking Experimental in Google AI Studio earlier this year, combining the speed of Flash with the capacity to solve increasingly challenging issues to enhance performance.
Additionally, it released an upgraded 2.0 Flash last week for all desktop and mobile Gemini app users, enabling them to explore new ways to create, engage, and work together with Gemini.
It is releasing the improved Gemini 2.0 Flash to the public today through Vertex AI and Google AI Studio’s Gemini API. Developers can now use Flash 2.0 to create production applications.
Additionally, it is releasing an experimental version of Gemini 2.0 Pro, which is its greatest model to date for complicated prompts and coding performance. It is accessible through Vertex AI, Google AI Studio, and the Gemini app for Gemini Advanced users.
Its most economical model to date, Gemini 2.0 Flash-Lite, is being made available to the public in preview form in Google AI Studio and Vertex AI.
Lastly, users of the Gemini app will be able to access 2.0 Flash Thinking Experimental through the desktop and mobile model selection.
Multimodal input with text output will be available in all of these models at release, and additional modalities will be prepared for widespread use in the upcoming months. Additional details, including price details, are available on the Google for Developers blog. Google is developing further features and updates for the Gemini 2.0 family of products in the future.
2.0 Flash: a fresh upgrade that will be widely accessible
The Flash series of models was first shown at I/O 2024 and is well-liked by developers as a strong workhorse model that excels at multimodal reasoning over enormous volumes of data with a context window of one million tokens. It is also ideal for high-volume, high-frequency jobs at scale. The developer community’s response to it has delighted us.
In addition to better performance in important benchmarks, 2.0 Flash is now widely accessible to a wider audience across all of its AI products, with text-to-speech and image generation on the horizon.
Try the Gemini API in Google AI Studio and Vertex AI, or Gemini 2.0 Flash in the Gemini app. You may find pricing information on the Google for Developers blog.
Its greatest model to date for complex prompts and coding performance is 2.0 Pro Experimental.
It received great input from developers regarding Gemini 2.0’s finest use cases, such as coding, and its strengths as it has continued to distribute early, experimental versions, such as Gemini-Exp-1206.
In response to that input, it is launching an experimental version of Gemini 2.0 Pro today. It outperforms all of the models it has released to date in terms of coding performance, handling difficult prompts, and comprehending and reasoning about world knowledge. In addition to having its largest context window 2 million tokens it can execute code and use tools like Google Search, allowing it to thoroughly analyse and comprehend enormous volumes of data.
Developers using Google AI Studio and Vertex AI, as well as Gemini Advanced users using the model drop-down on desktop and mobile devices, can now access Gemini 2.0 Pro as an experimental model.
Gemini 2.0 Flash-Lite: Google’s most affordable model to date
The cost and speed of the 1.5 Flash have received a lot of favourable reviews. Its goal was to continue raising quality while keeping costs and speeds constant. Today, it is launching the Gemini 2.0 Flash-Lite, a new model that is faster and costs the same as the 1.5 Flash but has greater quality. It performs better on most benchmarks than 1.5 Flash.
It features multimodal input and a context window with one million tokens, just like 2.0 Flash. For instance, under Google AI Studio’s paid tier, it can produce a pertinent one-line description for around 40,000 distinct images for less than $1.
Vertex AI and Google AI Studio are offering a public preview of Gemini 2.0 Flash-Lite.
Build with Gemini 2.0 Flash and Flash-Lite
Since the Gemini 2.0 Flash model family was introduced, developers have been finding new applications for this incredibly effective model family. In addition to having better speed than 1.5 Flash and 1.5 Pro, Gemini 2.0 Flash has more straightforward pricing, which lowers the cost of its 1 million token context window.
Gemini 2.0 Flash-Lite is now widely accessible through the Gemini API for enterprise clients on Vertex AI and for production use in Google AI Studio. Compared to 1.5 Flash, 2.0 Flash-Lite performs better on benchmarks for reasoning, multimodal, math, and factuality. With streamlined pricing for prompts exceeding 128K tokens, 2.0 Flash-Lite is an even more affordable option for projects that need lengthy context windows.
The 2.0 Flash family’s speed, efficacy, and affordability are already being used by developers to create amazing apps. Here are few instances:
AI voice recognition
Accuracy and speed are essential for developing conversational AI, especially voice assistants. Along with the capacity to manage intricate instructions and communicate with other systems through function calling, a quick Time-to-First-Token (TTFT) is necessary to produce a responsive, natural feel.
Gemini 2.0 Flash-Lite is being used by Daily to assist developers in producing innovative speech AI experiences. Daily has developed a system instruction code demo to accurately identify voicemail systems and customise messages using their open-source, vendor-neutral Pipecat framework for voice and multimodal conversational agents.
Data Analytics
By offering rich, insightful insights enabled by Gemini 2.0 Flash, Dawn is transforming the way engineering teams keep an eye on their AI products while they are in production. Engineering teams can quickly search through vast streams of user interactions using Dawn’s “semantic monitoring” pipeline to find any behaviour they’re looking for, such as user frustration, conversation length, and feedback. They can then continuously track these interactions as ongoing issues or topics to find anomalies and hidden problems in production.
By switching models, Dawn was able to drastically cut search times (from hours to just under a minute), save more than 90% on costs, and observe improved reliability across evals and production monitoring thanks to Gemini 2.0 Flash’s straightforward pricing, dependable structured outputs, and expanded context capabilities.
Editing videos
Mosaic is using Gemini 2.0 Flash to create a new agentic paradigm that is revolutionising time-consuming and sophisticated video editing activities. Using multimodal editing agents and Gemini 2.0 Flash’s long-context capabilities, their solution reduces tedious video editing processes from hours to seconds. This allows you to do things like create YouTube Shorts from any segment of a long-form film at the touch of a button.
Huge context windows are now 33% more economical with Google AI Studio’s new, simplified price for Gemini 2.0 Flash, which is $0.10 per million input tokens. This creates new opportunities for AI-driven video editing workflows.
