Newsletter
When it comes to unveiling AI models and tools, Google struck out twice. Once with Bard’s mistake and once with the faked Gemini hands-on video. Despite that, the company is pressing on. According to a new report, Gemini is now available for devs to use starting today. This is just another way to access Gemini.
A large part of the AI experience is separate from chatbots. Sure, tools like ChatGPT and Bard are great, but the underlying technology powering them is well-sought-after by developers. Imagine having access to an extremely smart brain that you can use for whatever you want. This is why GPT-4 is in high demand. However, Gemini might have something that will make people want to use it over GPT-4.
Gemini is now available to devs
If you’re a developer looking to create your own AI tool, then you’re going to need access to AI models to power your ambitions. Meta is very vocal about providing open-source AI tools to the public. Having access to several of the tools now requires crawling over paywalls.
Well, Google might not have a tool that’s open-source, but it’s free to use. The company just released Gemini for developers to use for their own applications. There are SDKs available to help you develop apps that use the Gemini model. You’re able to use Python, Kotlin, Node.js, Swift, and JavaScript for your products. Other features include “function calling, embeddings, semantic retrieval and custom knowledge grounding and chat functionality.”
Now, Gemini is an extremely powerful model, but you’re not getting the most powerful version. Google is letting people use Gemini Pro, which is the middle child sitting between Gemini Nano (the smallest model) and Gemini Ultra (the most powerful model). Google says that this tool is mostly for independent developers. So, if you’re planning on using this for hardcore enterprise applications, you might see a bit of a bottleneck.
Capabilities
Google is rolling this out now, and it’s going to get more powerful as time goes on. Right now, this implementation of Gemini has a context window of 32K, but the company says that it will grow as time goes on. Also, the version of Gemini available can accept and output text only. However, there’s also a Gemini Vision endpoint. This will allow the model to accept images as input.
If you want to use Gemini Pro, again, it’s free. You can access it through Google AI Studio and Vertex AI. However, the free access isn’t going to be forever. At launch, people will have access to 60 requests per minute as a “free quota.”
Early next year, the company is going to start charging you $0.00025 for every 1,000 characters of input. So, you’ll pay 1¢ for every 4 million characters. When it comes to images, you’re looking at $0.0025 per image. So, you’ll pay 1¢ for every 400 images.
You’ll also pay for the outputs as well. This is $0.0005 for every 1,000 characters. This boils down to 1¢ for every 2 million characters output. This won’t be a huge deal for smaller applications.
Further expansion
Google looks like it’s only halfway done with its master plan for Gemini. The company is planning on releasing the tool to more platforms as time goes on. This includes Duet AI, Firebase, Codelab, and Flutter.