Speaker 1: If you do a Google search for Gemini, you're probably going to find a lot of things here. There is Gemini, the chatbot, and the mobile app, and well, you can skip past the astrology and the crypto exchange, and then you find also the models, the chatbot, and Gemini for Google Workspace. And if you search for Gemini API, you find a Gemini developer API, but also a Gemini API that lives in Google Cloud. So it can be a little bit confusing. So in this video, I'll break down exactly what are the differences between Google AI Studio and the Gemini API for developers, and Gemini and Vertex AI, and what makes it an enterprise-grade platform. I'll also talk about Gemini, the chatbot, the difference in pricing and capabilities. So let's dive in. So first, Google AI Studio, here's where you can get an API key, and then you can configure that with one of the client libraries for the Gemini API for developers, like the Google Generative AI package in Python. And with a few lines of code, as you see here, you can import the library, set the API key, and then start prompting the models. And Google AI Studio is really the interface on top of that API that allows you to interact with it in a much more user-friendly way. And just to keep things a little bit more confusing, that API happens to be tied to a GCP project. So GCP is the Google Cloud Platform. It's basically the place where you can find infrastructure services, things like Compute Engine, Cloud SQL, and databases, and other stuff. And in that platform, you also have something called Vertex AI, which is an AI ML platform where you have the Gemini API in there, but also things like training and fine-tuning capabilities, ML Ops and governance tools, and a lot more. So Vertex AI is an end-to-end machine learning platform that lives in the Google Cloud Platform. And then within Google Cloud, we also have Google Workspace, where we have Docs, Slides, Sheets, and also Gemini, which has a chatbot interface, and a plan called Gemini Advanced, which allows you to connect to your Google Workspace files and content and things like YouTube. I'll get back to that later. Next, let's talk about Google AI for Developers, where you can find the Gemini API and AI Studio. It's where you can interact with Gemini easily with an API key. And this lives in the Google AI for Developers world. And in here, you can access the AI Studio interface. So let's take a look here. So the first thing you do is create an API key, accept the terms and a use policy. And here you can create an API key that is tied to a Google Cloud project with a billing account set up. The prices for Gemini 1.5 Pearl in the AI Studio are the ones you see here. There's a free tier with two requests per minute and 32,000 tokens per minute. And there's a price of $3.5 per 1 million tokens or 7 if the prompt is longer than 128K. For Gemini 1.5 Flash, the prices are a lot cheaper, as you see here. And the free tier is also a bit more generous. So in the AI Studio main interface, you have here the ability to add system instructions, select the model, including experimental model endpoints, which are sort of better versions of newer models that you can play with. Then there are controls like temperature, things like enforcing a JSON mode output. And you can edit the response schema here as well, including a visual editor, where you can just specify in a more friendly interface what are the properties and the value types that you need in your JSON response. And then you can also enable things like code execution if you want Gemini to run code for you in the process of coming up with an answer. Then you have advanced generation settings here, things like the safety settings and how stringent you want to be with these filters here or not. And then things like stop sequence, output length, top key, and top key. Now one nice thing about this interface is that once you type your prompts in here, you can also go and click on the Get Code button up on the top right corner, and you get the SDK equivalent of not only the settings you have in the UI, but also all the input it put in there. And then you can copy it or open it in a Colab notebook. One useful capability here as well is to compare responses from two different models. So if you click on the Compare button, you can select a different model here on the right hand side. And you can type a prompt that will be sent to both models at the same time, and you can see their responses in here. And funnily enough, the experimental endpoint response here was interrupted by a safety filter problem. But anyway, you can also fine tune a model here by providing a dataset with examples, and there's a nice documentation here that explains how to format that dataset. It's basically a bunch of examples of input and output, and the recommendation is to do at least 100 of those. There is also a prompt gallery with a lot of examples that you can work with. If you need some inspiration, you can always, of course, edit those prompts to customize it to your liking. There is also a community forum where you can see some of the discussions people have here and you can write your own topics and questions and engage with Googlers, but also the wider community. Now let's talk about Vertex AI. So Vertex AI is the enterprise-grade AI ML platform in the Google Cloud Platform. You can see here that we are under the Google Cloud documentation page. So we're no longer talking about Gemini for developers or AI Studio. We are talking about the Google Cloud Platform. And a Gemini API happens to be one of the many capabilities available in here. So if we go over here to Gemini API Quick Start, we can see some code samples, and it's actually very similar to how it works with Gemini API for developers. You need to import the Vertex AI library, that's the difference, and initialize the Vertex AI client. But it's otherwise fairly similar. Now on the Google Cloud Platform, if you navigate to Vertex AI, that's where you can see all of the tooling and services and capabilities available, which includes also an AI Studio, which is called Vertex AI Studio. Here you can also interact with Gemini. It's actually a very similar interface to Google AI Studio. And here you can ask your prompts and have controls that are similar to the ones we saw in AI Studio. And by the way, on the Google Cloud Platform, you also have a Gemini chat pane, where you can ask questions about the platform itself, if you're someone looking to learn more about how to use it. Yeah, a lot of things are named Gemini in the Google ecosystem. Now you may be wondering what makes Vertex AI an enterprise-grade platform. And there is things like sensitive data protection service with masking and tokenization capabilities, or the ability to create perimeters to isolate resources and control limit access. Then there are things like access transparency, which is the guarantee that customer data is not accessed for any reason other than to fulfill contractor obligations. There is a suite of compliance offerings here to really help companies in regulated industries. Identity clauses in a contract, both in terms of training data protection, but also generated output. And that helps reduce legal risks. And then there are things like service perimeters, where you can control very granularly what can access what and protect your resources. There are things like data residency for, again, compliance and regulatory frameworks here that would require that. Now in terms of pricing, Gemini on Vertex AI costs a little bit different than the price we sell for a Google AI Studio. If we normalize for 1 million tokens, we can see that for 1.5 flash, for example, in AI Studio, we have 0.075 versus 0.018 in Vertex AI. Or for Gemini 1.5 Pro, we have the prices that you see here. So basically in Vertex AI, things are a little bit cheaper. And probably the reason is in Vertex AI, a lot more consumption is expected. Those are enterprise customers and companies of large scale. And I guess you could argue AI Studio charges are premium for the more seamless experience. And also there is the free tier that is probably accounted for. So what the exact reasons for the price differences are, I'm actually not sure myself. But well, these are the prices, at least at the time of recording. Now let's talk about Gemini, the chatbot, and specifically Gemini Advanced. Here you can ask any kind of questions you want. And because it's in the Google ecosystem, you have a couple of cool things you can do here. So first, just control responses, also export to Docs and Gmail, and also check the facts with Google Search. That will highlight some of the components of the responses, and then tell you which search results are backing up this response here. Okay, in terms of media input, there are a couple of options here. You can upload files or add from drive or images. You can also, if you enable the Google Extensions over in Settings and Extensions, you'll see things like Google Workspace, Google Flights, Hotels, and YouTube, and some others. And then you can access them by typing the add symbol and specifying the extension you want to use here to retrieve information, for example, from your Drive documents. This is, by the way, coming from this document over here. Or you can use the YouTube extension to search for videos and ask questions about their contents. So there you go. Pretty nice set of integrations, and in terms of pricing, there is a free version where you get access to the 1.5 Flash model and 32,000 token context window, and still some capabilities like generate images and access some of the other Google apps. Then there's the paid version, which costs $20 a month, where you get 1.5 Pro access, 1 million token context window, priority access for new features, the ability to run code, and two terabytes of storage from Google One. Also Gemini in Workspace products. And if you are an Android user, there's also, of course, the Gemini mobile app and the Gemini Assistant and other capabilities available in Pixel phones or Samsung phones. Okay, that's a lot of information I know, so let me try and summarize. AI Studio is the user interface for Gemini API for developers. It's easy to get started there. There is a free tier. It also has a community forum and other cool features like the ability to compare responses and other capabilities. Then there's Vertex AI as the AI ML platform within Google Cloud. It is an enterprise-grade, compliance-ready environment with a lot of security controls and privacy controls. And besides Generate API, there's a lot more you can do in terms of more traditional machine learning with ML Ops and governance capabilities in there as well. And there's also a user interface that happens to be called Vertex AI Studio with a very similar look and feel to Google AI Studio. They only have Gemini or gemini.google.com, which is Google's version of Chat GPT. It's really basically a chatbot with plugins and extensions and great user experience. And it is meant for users, not builders or developers. And finally, the Gemini chatbot includes a free and a paid plan as well.