⚡ Introducing Gemini 1.5: Google's Latest AI Leap
posted 15 Feb 2024
Google has launched Gemini 1.5, the latest version of its language model. Initially targeting corporate users and developers, Gemini 1.5 is set to eventually become accessible to the broader public. The new model significantly outperforms its previous version, showing almost a 100% improvement in tests.
Gemini 1.5 Pro's performance, notably 87% better than its predecessor, verges on matching the prowess of Gemini Ultra. This enhancement stems from the 'Mixture of Experts' machine learning technique, which merges several smaller models into a larger, more efficient one without activating each component for every individual query.
Another key advancement is the increased token capacity of Gemini 1.5, now at 1 million tokens, vastly expanding its ability to process larger queries. In comparison, GPT-4 offers 128,000 tokens, and the standard Gemini Pro offers 32,000. Google CEO Sundar Pichai equates this capacity to processing approximately 11 hours of video.
I view it as one of the bigger breakthroughs we have done,Pichai remarked.
While Google aims to further augment the token capacity to 10 million, currently, usage of 1 million tokens is subject to additional fees. Nevertheless, Google's free applications will solely utilize Gemini 1.5 Pro with a 125,000 token limit, a change that Pichai assures will not adversely affect regular users.
More breaking news
