Google Unveils Gemini 1.5 Pro With 1 Million Token Context Window

Google announced Gemini 1.5 Pro on February 15, 2024, a new generation of its flagship AI model that includes a breakthrough 1 million token context window. This expanded context window represents a 100x increase over the previous 100,000 token limit of Gemini 1.0 Pro, enabling the model to process and reason over significantly larger amounts of information. The company stated that this capability allows Gemini 1.5 Pro to analyze up to 1 hour of video, 11 hours of audio, or over 30,000 lines of code in a single prompt.

During a demonstration, Google showcased Gemini 1.5 Pro's ability to analyze a 44-minute silent film, identifying specific objects and events within the video. The model also demonstrated its capacity to sift through a large codebase, pinpointing specific functions and potential bugs. This enhanced context window is powered by a new Mixture-of-Experts (MoE) architecture, which Google claims makes the model more efficient and performant. The MoE architecture allows the model to selectively activate relevant parts of its neural network for specific tasks, rather than engaging the entire model.

Gemini 1.5 Pro is currently available in a limited preview for developers and enterprise customers, with plans for broader availability in the future. Google emphasized that the model maintains strong performance across various modalities, including text, image, audio, and video. The company also highlighted its commitment to responsible AI development, noting that safety and ethical considerations have been integrated into the model's training and deployment processes. The increased context window is expected to unlock new applications in areas such as complex code analysis, long-document summarization, and detailed video content understanding.