As of: 2026-10-10 23:43 CEST.
Researched and written by AI agents (Claude Opus 5.5 in Claude Code). Human editor: Adam Bicz.
Canonical page: https://postcutoff.com/m/nova-2-5-sonic/

# Amazon Nova 2.5 Sonic

Amazon, Nova 2 family. Status: current. Type: speech and audio. Released: 2026-10-05. Training cutoff: not published.

## How to call it

| Provider | Model id | Endpoint | Docs |
|---|---|---|---|
| AWS Bedrock |  | https://bedrock-runtime.{region}.amazonaws.com (bidirectional streaming) | https://aws.amazon.com/about-aws/whats-new/2026/10/amazon-nova-2.5-sonic/ |

## Pricing

- Price not published.
- Unit as published: Same pricing as Nova 2 Sonic (AWS); see nova-2-sonic for figures
- Source: https://aws.amazon.com/about-aws/whats-new/2026/10/amazon-nova-2.5-sonic/
- Last checked: 2026-10-10

## Specs

- Input: audio, text
- Output: audio, text
- Context window: 256,000 tokens
- Max output: not published
- Open weights: no
- Licence: proprietary

## What it has not seen

- No published training cutoff. Counted from its release (2026-10-05): at least 143 AI events since (27 major, 7 historic), as of 2026-10-10.
- Everything since September 2026: https://postcutoff.com/since/2026-09/

## What stands out

- Asynchronous tool calling in voice: Voice agents can trigger tool calls without stalling the conversation; voice and text in the same session; 256K context; works with Strands Bidi Agents (AWS). Source: https://aws.amazon.com/about-aws/whats-new/2026/10/amazon-nova-2.5-sonic/

## Notes

GA Oct 5, 2026 in US East (N. Virginia), US West (Oregon), Europe (Stockholm), Asia Pacific (Tokyo); seven languages. Improved reasoning, instruction following and tool-calling accuracy and lower latency vs Nova 2 Sonic (AWS). Bedrock model id not verified (third-party sites list 'nova-2-5-sonic').

Successor to `nova-2-sonic`, Amazon's real-time speech-to-speech model for voice agents.

Model page: https://postcutoff.com/m/nova-2-5-sonic/