Announcing Amazon Nova 2.5 Sonic with improved reasoning for voice agents

Today, we announce the general availability of Amazon Nova 2.5 Sonic, our latest speech-to-speech model for natural, real-time voice agents. This release brings improved reasoning, instruction following, and tool-calling accuracy to real-time voice applications. The model also delivers lower latency, enabling faster, more natural interactions.

Nova 2.5 Sonic helps voice agents maintain context, follow multi-step instructions, and use connected tools to complete customer tasks. For example, a customer-service agent can look up an order, check return eligibility, and initiate a return within a conversation. These improvements help developers build more capable voice agents for customer service, personal AI assistants, and real-time conversational applications.

Nova 2.5 Sonic builds on existing capabilities, including robust speech understanding, expressive voices in seven languages, and controllable turn-taking, alongside support for voice and text in the same session, asynchronous tool calling, and a 256K context window. Developers can also use Nova 2.5 Sonic with Strands Bidi Agents, now generally available, to build production voice agents in a few lines of code, with longer conversations and connections to other agents.

Nova 2.5 Sonic is available in Amazon Bedrock in the US East (N. Virginia), US West (Oregon), Europe (Stockholm), and Asia Pacific (Tokyo) AWS Regions, at the same pricing as Nova 2 Sonic. To learn more, visit the Amazon Nova models page. To get started, see the Amazon Nova user guide, and blog on building voice agents with Nova 2.5 Sonic and Strands Bidi Agents.