Amazon Nova Sonic
Speech-to-speech foundation model that listens, reasons and speaks in one model for real-time voice agents on Bedrock.
About
Amazon Nova Sonic is Amazon's speech-to-speech foundation model for real-time voice agents. Instead of chaining a speech recognizer, a language model and a text-to-speech system, it listens, reasons and speaks inside a single model, which cuts the latency that usually makes chained voice pipelines feel sluggish. It runs on Amazon Bedrock through a bidirectional streaming API, takes speech and text in the same session, and can call tools or pull in knowledge mid-conversation, so an agent can look something up without dropping the thread.
Amazon announced it on April 8, 2025, positioning it as a cheaper path than separate ASR, LLM and TTS services and as a rival to OpenAI's Realtime API and Google's native audio models. It launched with a 300,000-token context window and American and British English, later adding Spanish, French, Italian and German, with a second generation (Nova 2 Sonic) following in December 2025 and Nova 2.5 Sonic in October 2026. Billing is pay-as-you-go on Bedrock with text and speech tokens metered separately.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
No independent published benchmarks; Amazon's 'industry-leading price performance' and latency claims are provider marketing (one AWS-partner case study cites ~$0.27/hour of input audio in production)
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 8/10
Scored 8/10 (TIER verified, metrics-v1). Evidence: AWS announcement (April 8, 2025) and Bedrock product page; bidirectional streaming API live on Bedrock; 300K context window per provider docs. Strengths: unified model removes chained-pipeline latency, tool use and knowledge grounding built in, backed by AWS enterprise support. Docked for: (1) no independent benchmarks — performance claims are provider marketing; (2) closed weights, Bedrock-only, and pricing that requires care to model (separate text/speech metering). License: proprietary (AWS service).
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 8.0/10
- Maintenance · recency, releases, issue load
- 8.0/10
- Openness · license + docs
- 3.0/10
- Manual review · sanity read + red-flag check
- 8.0/10
- · Announced April 8, 2025; Bedrock bidirectional streaming API
- · 300K context window (provider-announced); tool use + knowledge grounding
- · Nova 2 Sonic (Dec 2025) / 2.5 Sonic (Oct 2026) followed; original remains available
- · Pricing: text and speech tokens metered separately; AWS logo staged