Google Gemini 1.5 Pro was introduced by the tech giant. Google, one of the giant technology names, has launched Gemini 1.5, the new version of its large language model Gemini, for developers and corporate users.
The company states that with the new version, Gemini has made significant improvements in many ways and has made great progress.
What is Google Gemini 1.5 Pro?
Google notes substantial improvements achieved with Gemini 1.5. The Pro version aligns with the capabilities of the recently introduced high-end Gemini Ultra model. It also surpasses the earlier Gemini 1.0 Pro release by a notable 87% in benchmark assessments.
This performance increase is largely attributed to the use of “Mixture of Experts” (MoE) techniques. MoE allows the model to selectively activate specific sections during query processing rather than the whole model at once. This strategy promotes a faster user experience and greater efficiency on Google’s infrastructure.
Expanded contextual understanding
A substantial upgrade in Gemini 1.5 relates to handling significantly larger datasets for a single query. It features a context window of one million tokens, increasing from 128,000 in OpenAI’s GPT-4 to 32,000 in the previous Gemini Pro. This context window effectively permits Gemini 1.5 to process over 700,000 words, an hour-long video, eleven hours of audio recordings, or 30,000+ code snippets simultaneously. Google researchers are further investigating potential use cases exceeding even the one million token context.
Business applications
Google executives envision the expansive context window as particularly valuable for various business settings. It opens up applications where extensive personal data and information can be analyzed during queries. A hypothetical example highlights a filmmaker’s potential to submit an entire movie, querying Gemini 1.5 on anticipated critical reception. Financial sectors could also leverage Gemini 1.5’s increased capacity to scrutinize voluminous records and data sets.
Availability and accessibility
Initial access to Gemini 1.5 is intended for developers and commercial users through platforms like Google‘s Vertex AI and AI Studio. Subsequently, this new iteration will replace Gemini 1.0. The Gemini Pro version, offered publicly on gemini.google.com and across applications, will be Gemini 1.5 Pro but with a limited context window of 128,000 tokens. Developers and users desiring the maximum context window of one million tokens would incur additional costs. Google underscores ongoing tests to assess the model’s security and ethical boundaries, particularly surrounding the enhanced contextual capabilities.
Featured image credit: Google
