After introducing Gemini Advanced to the public, Google has announced its next-gen AI model called Gemini 1.5. The company says that the latest version of Gemini is leaps ahead in terms of performance, capable of handling longer sessions of coding, text summary, images, and more. Gemini 1.5 is a mid-size multimodal model that sits in between Gemini 1.0 Pro and Gemini 1.0 Ultra.
What is Gemini 1.5 Model by Google?
Gemini 1.5 introduces a new ‘Mixture-of-Experts’ architecture, which makes the AI model more efficient. In simple words, it is designed and trained for long-context understanding to help with practical tasks such as coding, image processing, and other complex calculations.
Google says that Gemini 1.5 Pro can handle up to 1 million tokens. Hence the new model will be able to remember more information in its memory, while pursuing further tasks. According to the company, Gemini 1.5 Pro is almost as good as Gemini 1.0 Ultra.
One of the major updates with this new model is the refined safety rules. Google mentions that they have been constantly evaluating risks related to powerful AI models, and have introduced new filters to prevent potential harms.
Gemini 1.5: Availability
Gemini 1.5 Pro is currently available to enterprise users of Google AI Studio and Vertex AI. The company has not disclosed a public rollout yet. For now, regular users can redeem the free trial of the Gemini Advanced chatbot which is capable of tasks like coding, image generation, and even replaces the Google Assistant on Android smartphones.
Gemini App comes to India
Google has also started the official rollout of the Gemini app in India. Earlier, the app was only available in the United States. The company is also pushing an update to the Google App on iOS, enabling the Gemini chatbot on iPhones and iPads.
A recent report revealed that Google is planning to bring Gemini to headphones. It could enable AI features like translation in smart devices like the Pixel Buds, allowing users to communicate in foreign languages in real-time. However, the company has not officially announced this feature yet.
