Google has launched Gemini 1.0 AI, it will be available in three sizes
December 8, 2023

Google has launched Gemini 1.0 AI, it will be available in three sizes

Blog

Google Gemini 1.0 was announced a few months ago at the Google I/O 2023 event, and today it finally launched this model in three sizes that will start arriving on the Bard. The company revealed a lot of details about how the new model will work, and shared some really impressive information.

Google claims to be Gemini 1.0 the most powerful general model available. A model can train separate components for different modalities and then merge them. It is important to note that the company stated that the model will have difficulty with anything that is more conceptual and requires more complex reasoning.

As for Gemini, Google has “pre-trained it from scratch on different modalities” using TPU 4 and TPU v5e. Google is yesterday announced in his blog and TPU v5p and claims to be “the most powerful, efficient and scalable“.

To show just how powerful Gemini really is, Google demonstrated how Gemini processes 200,000 scientific papers, filters the relevant ones, and then summarizes the data in an hour. Programming is another thing that Google paid attention to, and it was noted that the model will be able to understand, explain, and generate high-quality code in multiple languages ​​such as Python, Java, C++, and Go.

At the moment, the Gemini 1.0 is available in three different sizes:

  • Ultra – the largest and most powerful model for extremely complex tasks.
  • Pro – the best model for scaling across a wide range of tasks.
  • Nano – the most efficient model for on-device tasks.

Gemini 1.0 also has a lot of power in terms of multimodality. The Gemini Ultra beats the ChatGPT-4V in the image, video and audio tests.

Google Gemini is better than ChatGPT 3.5

Google also commented that it is one of the most secure AI models to date, so users who are concerned about security need not worry about it. Users will be able to experience the Pro version with Google Bardom, which means it will offer advanced reasoning, planning, and writing to help you understand and summarize content. The company also stated that the new model will be able to deliver better performance than ChatGPT 3.5.

Bard with Pro version will now be available in English for 170 countries, GeminiPro will only run text queries at first, but other modalities will be available soon. The Ultra version will be available next year, and Google is currently conducting “extensive trust and security checks” so that it can begin rolling out to all users.

All this really looks impressive, at least in words, and we can't wait to see how this AI model will develop over time, and of course, fight with the competition.