Google’s intention to capitalize on the AI wave resulted in Bard, later rebranded as Gemini.
The company announced improvements to the Gemini AI Pro model, released in February 2024.
In addition, Google announced the Gemini 1.5 Flash model, a lightweight LLM compared to the Gemini 1.5 Pro.
Two Gemma models, PaliGemma and Gemma 2, also debuted alongside the new Gemini models.
What’s Improved in Gemini 1.5 Pro?
Google shared on its official announcements page about the improvements to the Gemini 1.5 Pro model.
It is now better at translation, coding, reasoning, and more tasks.
The company assures you will see these updates in the model starting today. So, Gemini 1.5 Pro will produce better responses to your prompts with improved accuracy.
What’s Gemini 1.5 Flash?
Gemini 1.5 Flash is a lighter model trained by 1.5 Pro through “distillation.” Google says, “In this process, the most essential knowledge and skills from a larger model are transferred to a smaller, more efficient model.”
This new model is best suited for narrower or high-frequency tasks which demand a prompt response. So, it will better appeal to use cases such as chatbots.
Both of the new models support a 1 million token context window. A context window refers to the input size an LLM can process. So, you can input longer text chunks and bigger video/image files into the LLM.
The great news is that Google is doubling the size of the token context window to 2 million for Gemini 1.5 Pro. But you have to sign up for a waitlist.
How to Access These Models?
Please note that all these announced models are for developer use only.
Developers will access them via APIs and then create or improve their apps. Moreover, these models are available in Preview and will begin rolling out in June 2024.
Google also plans to introduce context caching in Gemini 1.5 Pro in the coming months. It will save time and cost of sending parts of a prompt, including large files, to the model. Android 15 will integrate Gemini to help respond to app-specific queries.
You can submit your apps to the Gemini API Developer Competition.
The best app developer will get a custom electric DeLorean (the model from Back to the Future).
