Google has announced the open beta testing of Gemini 1.5 Pro, an advanced neural network with a standout capability: handling up to one million tokens in a single session, according to reports from Analytics India Magazine.
The Gemini 1.5 Pro model features a context window that starts at 128k tokens and can be expanded to an impressive one million tokens. It can process roughly 700 thousand words, about 30 thousand lines of code, one hour of video, or up to 11 hours of audio material. In testing, Google also demonstrated successful processing of as many as 10 million tokens.
Constructed on Transformer architecture with Mixture of Experts capabilities, Gemini 1.5 Pro shows strong performance across tasks such as historical data analysis. For example, it can analyze transcripts from significant events like the Apollo 11 mission. Beyond large-scale data analysis, the system excels at locating specific text fragments quickly and handling substantial blocks of code efficiently.
In the Needle In A Haystack testing scenario, the network achieved 99 percent accuracy in pinpointing precise data within extensive documents. Its adaptive learning demonstrated in the Single Book Machine Translation benchmark positions Gemini 1.5 Pro among the leading models in contemporary AI research.
At present, users can access Gemini 1.5 Pro through Google AI Studio. The system understands and can respond in Russian, though it is not available in Russia at this time. Information about regional availability is provided by online services and the provider’s announcements.
Earlier, GigaChat was described in coverage by social media outlets. These developments reflect ongoing efforts to push the boundaries of natural language processing, reasoning over large datasets, and efficient code interpretation within large-scale AI systems. [Analytics India Magazine] The Gemini line continues to attract attention for its ambitious token handling and real-time inference capabilities.