YandexGPT training

YandexGPT is used in SpeechSense dialogs for data analysis. To increase the analysis quality, e.g., assign semantics tags or perform audio recognition, YandexGPT models can be trained on your project data.

Starting September 1, 2025, YandexGPT training is enabled by default in new projects. You can disable it if you prefer to opt out. For projects created before September 1, 2025, you need to enable YandexGPT training manually.

Data collection starts once you enable training, and the YandexGPT model gets updated every three months based on the collected data. The result of model training will only be integrated into your project.

To enable or disable YandexGPT training, you need the Admin role in the space.

Enabling YandexGPT trainingEnabling YandexGPT training

To enable YandexGPT training for one or multiple projects:

1. Open the SpeechSense home page.
2. Go to the space of your choice.
3. Navigate to the Settings tab.
4. Under Tuning YandexGPT, click Customize.
5. Select one or more projects you want to enable YandexGPT training for, then click Save.

From now on, all new data you add to the project will be collected for YandexGPT training. Older data is not used for training.

Disabling YandexGPT trainingDisabling YandexGPT training

To disable YandexGPT training for one or multiple projects:

1. Open the SpeechSense home page.
2. Go to the space of your choice.
3. Navigate to the Settings tab.
4. Under Tuning YandexGPT, click Customize.
5. Deselect those projects you want to disable YandexGPT training for, then click Save.