The Dialog-RSN-1 model from PolyAI is available on AWS Marketplace and Amazon SageMaker JumpStart
PolyAI has made the Dialog-RSN-1 model for real-time voice conversation available on AWS Marketplace and Amazon SageMaker JumpStart. The model can be deployed with one click directly into the customer's own AWS account, with isolation via VPC and IAM.
PolyAI has made its Dialog-RSN-1 model, designed for real-time voice conversation with reasoning, available on AWS Marketplace and in the Amazon SageMaker JumpStart catalog. This is the first PolyAI model offered through AWS Marketplace. AWS customers can subscribe to the model and deploy it as a SageMaker Studio endpoint with one click, or script the deployment via the SageMaker Python SDK.
According to the company, the model processes call audio directly in a single pass — without a separate speech-to-text transcription step and without passing data between multiple models — and decides when and how to respond. The endpoint communicates using the OpenAI Realtime protocol, so an existing Realtime client can connect just by changing the base URL. Input audio must be 16-bit PCM mono at 16 or 24 kHz (G.711 or Opus formats must first be transcoded), connections last a maximum of 30 minutes, and the model currently supports only English.
The deployment runs exclusively within the customer's own AWS account — in their own VPC, under their own IAM roles, and with monitoring via Amazon CloudWatch, just like other workloads. According to PolyAI, this means call audio and transcripts remain within a boundary that the customer controls themselves, which is intended to help with data residency and auditability requirements. The customer manages orchestration, telephone integration, and business logic themselves.
Before deployment, SageMaker JumpStart offers preset optimization profiles — focused on lowest latency, highest throughput, lowest cost per request, or a balanced trade-off — and displays metrics such as P50 latency, time to first token, and throughput, so the customer can select an appropriate ml.* instance without manual benchmarking. The source article further mentions additional parts of the deployment process that are not available — for details, see the source article.
Why it matters
Companies that run phone support or contact centers gain the ability to deploy a ready-made model for processing live calls without having to build their own speech recognition and reasoning pipeline. Because the runtime operates within the customer's own AWS account, this simplifies meeting data residency and auditability requirements for sensitive voice data. For developers, this shortens the path from model selection to production deployment thanks to preset optimization profiles.
Two audiences, two different impacts
What this means
For individuals
For developers and ML engineers building voice AI agents, the availability of the model via SageMaker JumpStart means the ability to deploy a ready-made model with one click, without tuning containers, instance types, and benchmarking infrastructure.
For a business
Companies running a telephone contact center can deploy the model for real-time voice conversations directly within their own AWS account (own VPC, IAM roles, CloudWatch monitoring), which, according to the company, helps meet data residency and auditability requirements without having to build their own pipeline for speech recognition and reasoning.
ProcessesCheck the original
Event sources
only one source so far · 1 publisher, 0 independent. We count feeds from the same owner only once.