Google has announced its latest AI model, Gemini 1.5, which can handle up to 1 million tokens of input, equivalent to about 4 million characters or 800,000 words⁶. This is a huge leap from the previous models, which had a maximum context window of 200,000 tokens⁵.
Gemini 1.5 is the result of Google's relentless innovation in natural language processing (NLP), the branch of AI that deals with understanding and generating human language. Google claims that Gemini 1.5 can deliver "dramatically enhanced performance" across a wide range of tasks and modalities, such as text, images, audio, and video⁵.
One of the key features of Gemini 1.5 is its new Mixture-of-Experts (MoE) architecture, which improves efficiency by routing the input to a group of smaller "expert" neural networks, instead of using a single large network. This way, the model can produce faster and higher quality responses⁵.
Gemini 1.5 also introduces a breakthrough experimental feature in long-context understanding, which enables the model to process and reason from large amounts of information, such as entire PDFs, code repositories, or lengthy videos. This opens up new possibilities for developers and users, who can upload multiple files and ask questions, or query an entire codebase and get insights⁵.
Google has made Gemini 1.5 available for private preview in Google AI Studio and Vertex AI, its platforms for building and deploying AI applications. The model is accessible in 38 languages across 180+ countries and territories⁵. Google plans to roll out more models in the Gemini 1.5 line, including Gemini 1.5 Ultra, which will be its largest and most powerful model yet⁴.
Gemini 1.5 is part of Google's vision to make AI more helpful and accessible for everyone. The company says that it is committed to ensuring the safety and responsibility of its AI models, and that it will continue to work with the research community and stakeholders to advance the field of AI⁵.
Source: Conversation with Bing, 2/18/2024
(1) Meet Gemini 1.5: Google's latest weapon in the AI race. https://interestingengineering.com/innovation/googles-new-ai-model-is-700x-more-powerful-than-openais-flagship.
(2) Introducing Gemini 1.5, Google's next-generation AI model. https://blog.google/technology/ai/google-gemini-next-generation-model-february-2024/.
(3) Gemini 1.5: Our next-generation model, now available for Private .... https://developers.googleblog.com/2024/02/gemini-15-available-for-private-preview-in-google-ai-studio.html.
(4) Gemini Pro 1.5 with 1 million tokens surpasses GPT-4 Turbo: What does that mean?. https://indianexpress.com/article/explained/explained-sci-tech/google-gemini-pro-1-5-1-million-tokens-9166398/.
(5) Know which OS supports Google Gemini 1.5 now available in India. https://odishatv.in/news/technology/know-which-os-supports-google-gemini-1-5-now-available-in-india-228077.
(6) Google unveils Gemini 1.5 AI model with enhanced performance, long context window & more. https://www.digit.in/news/general/google-unveils-gemini-15-ai-model-with-enhanced-performance-long-context-window-more.html.
