Skip to main content
Core SDK 0.26.0 downloads Quail Multi Speaker, Rook Multi Speaker and VAD Multi Speaker only by their new IDs. This page covers the model ID change only. To find the core SDK version of your package, use the compatibility matrix.

​
Model file version 8: use the new IDs

Core SDK 0.26.0 requires model file version 8. Version 8 files exist only under the new model IDs from the September 2026 rename. The old IDs have no version 8 files. Core SDK 0.25.2 still downloads and loads them as version 7 files. Before you upgrade to core SDK 0.26.0:
  1. Replace each old ID in your code and configuration with the new ID from the table.
  2. Download the version 8 model file for the new ID. Use the model download function of your SDK or artifacts.ai-coustics.io.
  3. Remove version 7 model files from your deployment. Core SDK 0.26.0 rejects them with AIC_ERROR_CODE_MODEL_VERSION_UNSUPPORTED.
If you keep an old ID, the model download fails because no version 8 file exists for it. The Python, Node.js and Rust SDKs report this error:
You can change to the new IDs before you upgrade. Core SDK 0.25.2 also downloads the new IDs as version 7 files.

​
Quail -> Quail Multi Speaker

In September 2026, Quail became Quail Multi Speaker to distinguish the general-purpose model from Quail Voice Focus. Quail Voice Focus was not part of this naming change and retains the quail-vf-* naming pattern. This does not mean every version remains supported; see deprecated models for removed Voice Focus IDs.

​
VAD: Quail VAD -> VAD -> VAD Multi Speaker

The voice activity detection (VAD) models dropped the Quail prefix in August 2026. In September, the general-purpose model gained an -ms- infix to distinguish VAD Multi Speaker from VAD Voice Focus. Current rename VAD Voice Focus was not part of this step and keeps its vad-vf-2.0-s-16khz ID. August transition: these intermediate names precede the September rename above.
VAD moved from 2.0 to 2.1 in this step, so it was both a rename and a newer model version. VAD Voice Focus stayed at 2.0.

​
Rook -> Rook Multi Speaker

In September 2026, the perceptual speech enhancement models became Rook Multi Speaker. Their IDs gained an -ms- infix.
Some old model IDs on this page are no longer supported by the current core SDK. Update to the current IDs above. For models that have been removed, see deprecated models for their last supported core SDK versions.