Google Unveils Gemini 2.5 Ultra with 10M Context Window and Native Code Execution

In a major leap forward for enterprise AI infrastructure, **Google** has officially launched **Gemini 2.5 Ultra**, introducing a record-breaking 10-million token context window alongside deep native code execution capabilities. The model represents Google's answer to growing developer demand for long-context reasoning across massive codebases and enterprise document repositories.

Built directly into Google Cloud Vertex AI and Gemini API endpoints, Gemini 2.5 Ultra features an integrated Python execution environment that allows the model to run, test, and debug code snippets within a micro-sandbox in real-time. Benchmark evaluations show a 40% reduction in end-to-end inference latency compared to Gemini 1.5 Pro.

Google also unveiled updated pricing tiers that reduce long-context input token costs by 50%, making full-repository analysis economically viable for software development teams and data science workflows worldwide.