After announcing its family of Gemini models last week and bringing it to its Bard chatbot experience, Google is now bringing Gemini to developers by launching a slew of new and updated services today. One of these services is AI Studio — which was previously known as MakerSuite.
AI Studio is a web-based tool for developers that functions a bit like a gateway into the wider Gemini ecosystem, starting with Gemini Pro and then, at some point next year, also Gemini Ultra. Using the service, developers can quickly develop prompts and Gemini-based chatbots — and then get API keys to use them in their apps or get access to the code to work on it in a more fully featured IDE.
It’s important to note that there is a relatively generous free quota, with up to 60 requests per second, which should be enough to quickly iterate on ideas without facing onerous restrictions and maybe even enough to power some lesser-used applications in productions.
There is a price to pay here though: For developers using the free tier (and that’s pretty much everyone for now, as Google only plans to launch a paid version early next year — likely to coincide with the GA launch of the Gemini Ultra model), Google’s reviewers can see the input and output of the API and web app to “improve product quality.” Google notes that this data is de-identified from the user’s Google Account and API key, though.
Compared to the earlier version of MakerSuite/AI Studio, this updated edition feels quite a bit more substantial. Among other things, it will offer support for both Gemini Pro and the Gemini Pro Vision model, allowing developers to work with both text and imagery (though not for creating images).
“We’ve designed it really to be the fastest way to build with Gemini,” Josh Woodward, Google’s VP for Google Labs, told me. “We really want to invite developers to come play with it. It is the first version and we’ve got a lot of fine-tuning we’re already doing now for future updates, too, but we’re trying to design it in a way where people can just get in and really start building with it.”
In the web interface, developers can choose their models, adjust the temperature to control the output’s creative range and provide examples to provide tone and style instructions. You can also tune the model’s safety settings. It’s worth noting that in MakerSuite, you could turn off all the guardrails by telling the system not to block any responses, while in AI Studio, the lowest setting now is “block few.”
There are also different workflows for creating freeform, structured and chat prompts, too.
Woodward noted that the team tried to design AI Studio so even the free tier wouldn’t feel like a trial or gated product. Indeed, assuming the free tier’s rate limits are sufficient for their use cases, developers can start publishing their AI Studio applications or use them through the API or Google’s SDKs right away.
Jeanine Banks, the VP and GM for Google’s Developer X teams and head of developer relations, also stressed that AI Studio is a gateway into Google’s wider AI ecosystem and in particular to Vertex AI, Google’s enterprise-ready generative AI developer platform.
She noted that the team wants this transition between going from AI Studio and to Vertex to be as seamless as possible. Woodward added that the strong SDK support came directly from user feedback. “The first version we showed people, they said: ‘I love how easy it is to prompt. Now I want to go to code.’ And there was sort of this cliff that we had to fill in,” he told me.
Talking about the overall ecosystem, Banks also explained that Google plans to bring Gemini to the Chrome Dev Tools and Google’s Firebase mobile development platform early next year.
Given the speed at which generative AI is developing, it’s hard to even predict what developers will want to use these tools for next, but Banks and Woodward stressed that Google plans to build out AI Studio as an easy onramp for developers of all skill levels.
“I hope that AI Studio, in some ways, won’t just be seen as a prompting tool or something that only developers go to, but it’s actually, in some ways, a developer and creativity tool, where people can come up with ideas for working with these models and all the capabilities that will come out over the next year or so,” Woodward said.