Amazon Nova 2.5 Sonic: AWS Introduces Advanced Voice Agent AI Model
AWS has announced the general availability of Amazon Nova 2.5 Sonic, a speech-to-speech AI model built for real-time voice agents. The model delivers improved reasoning, instruction following, and tool calling accuracy for natural, low-latency voice conversations. Enterprises and developers can use it for customer service, personal AI assistants, and real-time conversational applications.
Amazon Nova 2.5 Sonic: AWS Introduces Advanced Voice Agent AI Model
Amazon Web Services (AWS) has announced the general availability of Amazon Nova 2.5 Sonic, its latest speech-to-speech AI model designed for real-time voice agents. The model was unveiled on October 5, 2026, and brings improved reasoning, instruction following, and tool calling accuracy to natural, conversational voice applications.
What Is Nova 2.5 Sonic?
Amazon Nova 2.5 Sonic is the newest model in the Amazon Nova family, built specifically for voice AI. Unlike traditional text-based AI systems, it processes speech-to-speech conversations in real time, allowing users to interact with AI through natural voice dialogue. The model supports both voice and text inputs within the same session, enabling seamless transitions between speaking and typing.
Improved Reasoning and Tool Calling
The key improvement in Nova 2.5 Sonic lies in its ability to maintain context and follow multi-step instructions. In practical terms, a customer-service agent powered by the model can conduct a conversation, look up an order, check return eligibility, and initiate a return all within the same dialogue. The model also supports asynchronous tool calling, meaning it can continue running background tasks while the user continues the conversation.
Language and Voice Support
Nova 2.5 Sonic supports speech understanding in multiple languages and generates speech in eighteen expressive voices. These include masculine and feminine voices in English (US, British, Indian, and Australian variants), as well as Spanish, German, French, Italian, Portuguese (Brazilian), and Hindi. The model features controllable turn-taking, allowing it to pause, hesitate, or interrupt naturally during a conversation.
Availability and Pricing
The model is available through Amazon Bedrock in the US East (N. Virginia), US West (Oregon), Europe (Stockholm), and Asia Pacific (Tokyo) AWS Regions. Pricing is the same as Nova 2 Sonic. Developers can also pair Nova 2.5 Sonic with Strands Bidi Agents, which is now generally available, to build production voice agents with longer conversations and connections to other agents.
What This Means for Businesses
Voice agents are becoming increasingly important for customer service, education, and language learning. With lower latency and improved reasoning, Nova 2.5 Sonic enables more natural interactions between AI and users. Businesses in travel, telecommunications, entertainment, and other industries can use the model to automate customer support, build personal AI assistants, and create real-time conversational applications.
Sources:
- AWS News Blog: Announcing Amazon Nova 2.5 Sonic with improved reasoning for voice agents (https://aws.amazon.com/about-aws/whats-new/2026/10/amazon-nova-2.5-Sonic/)
- AWS Service Card: Amazon Nova 2.5 Sonic (https://docs.aws.amazon.com/ai/responsible-ai/nova-2-5-sonic/overview.html)
- AWS Nova Models Page: Amazon Nova (https://aws.amazon.com/nova/models/)
What's Your Reaction?
Like
0
Dislike
0
Love
0
Funny
0
Angry
0
Sad
0
Wow
0