Google has published a model card for Gemini 2.0 Flash, a natively multimodal model in the Gemini 2.0 series designed to power agentic systems. The model improves upon Gemini 1.5 Flash with enhanced quality and outperforms Gemini 1.5 Pro on key benchmarks at twice the speed.
- Supports text, images, audio, and video inputs with a 1,048,576 token context window.
- Achieves 77.6% on MMLU-Pro, 34.5% on LiveCodeBench (v5), and 90.9% on MATH benchmarks.
- Delivers low-latency support for real-time streaming via the Multimodal Live API.
- Trained using Google's Tensor Processing Units (TPUs) with JAX and ML Pathways.
The model is positioned as an upgrade path for users seeking enhanced quality or slightly better performance with real-time latency capabilities.