Google Unveils Gemini 1.5 ai Model with Million-token Context Window
Google has introduced its latest artificial intelligence (ai) model, Gemini 1.5, featuring an “experimental” one million token context window. This new capability enables the ai to process extremely long text passages of up to one million characters in order to understand context and meaning.
Improved Performance
Gemini 1.5 Pro achieves near-perfect recall on long-context retrieval tasks across modalities, improving the state of the art in long-document question answering (QA), long-Website video integration QA, and long-context automatic speech recognition (ASR). It also matches or surpasses the performance of its predecessor, Gemini 1.0 Ultra.
Innovative Mixture-of-Experts (MoE) Architecture
The efficiency of Google’s latest ai model is attributed to its innovative Mixture-of-Experts (MoE) architecture. Unlike traditional Transformer models, which function as a single neural network, MoE models are divided into smaller “expert” neural networks that can be selectively activated depending on the input.
Powerful Demonstration
Google demonstrated the power of the one million token context window by ingesting the entire 326,914-token Apollo 11 flight transcript and accurately answering specific questions about it. It also summarized key details from a 684,000-token silent film when prompted.
Free Access for Developers and Enterprises
Initially, Google is providing developers and enterprises with free access to a limited Gemini 1.5 preview containing a one million token context window. A 128,000 token general release for the public will be available later, along with pricing details.
Potential Standard for ai Text Understanding
For now, the one million token capability remains experimental. However, if Gemini 1.5 lives up to its early promise, it could set a new standard for ai’s ability to understand complex, real–world text.
Testing Opportunities
Developers interested in testing Gemini 1.5 Pro can in ai Studio. Google says that enterprise customers can reach out to their Vertex ai account team.
