Nova 2.0 Pro Preview (Non-reasoning) is a high-speed multimodal foundation model developed by Amazon, introduced as part of the second-generation Nova model family in December 2025. This variant is specifically optimized for tasks that prioritize rapid text generation and immediate response over the extended thinking or reasoning phases found in reasoning-focused models. It supports both text and image inputs, providing a balance of performance and efficiency for real-time applications.
Technically, the model is characterized by high throughput, with reported median output speeds of approximately 154 tokens per second and low initial latency. It features a context window of 256,000 tokens, allowing it to process large documents or extensive conversational histories. Evaluations have noted its performance as being above average for non-reasoning models in its price tier, though it exhibits higher verbosity in its generated responses compared to its predecessors.
The model is designed for enterprise use cases such as content creation, summarization, and data extraction from multimodal sources. It serves as a cost-effective solution for developers who require high intelligence and speed for standard generative tasks without the computational overhead associated with the complex multi-step logical inference found in models with an active thinking budget.