AI models in SourceCraft

SourceCraft offers a wide range of AI models. They are used in the SourceCraft UI chat, AI coworkers, Code Assistant IDE plugins, and SourceCraft CLI.

Note

Model selection in VibeCraft will become available later.

Tip

We are constantly updating the list of models. You can view the current list and descriptions of models on the AI models page of your personal organization. To view, log in or sign up to SourceCraft. You can view another organization's models in the organization settings.

Default model

The SourceCraft Default model is an ensemble of models managed by SourceCraft. SourceCraft automatically selects a model to solve your task. This mode allows for operation without having to select a model manually.

Internal and external models

  • Internal models are deployed within the Yandex Cloud infrastructure. Data sent to the model does not leave the Yandex Cloud infrastructure.

    The infrastructure is protected in accordance with the requirements of the Russian Federal Law No. 152-FZ on Personal Data.

    Therefore, you may use internal models to process personal data in compliance with the requirements of Federal Law No. 152-FZ, as indicated by a specific note on the AI Models page.

    You can use the organization's gift quota of neurocredits for internal models.

  • External models are deployed outside the Yandex Cloud infrastructure.

    When selecting such a model, the data required to process the request is transmitted to an external provider. External models are not suitable for tasks requiring compliance with Federal Law No. 152-FZ. The organization's gift quota is unavailable for them.

Access to models

Note

To manage models, you need the organization Administrator or Owner role. They determine which models may be used by the organization's members. For example, you can leave only the internal models accessible. The setting is available regardless of the administrator's or organization owner's subscription.

For more information, see Configuring AI models in an organization.

Model switching becomes available to the user if the following conditions are met:

  • A paid subscription has been set up for the user within the organization they selected for neurocredit deductions.
  • An administrator or organization owner has configured the use of models other than the default model.

If a user has subscriptions assigned across multiple organizations, they can choose which organization to deduct neurocredits from. Access to models is determined by the subscription and the settings of the selected organization.

Note

Without a paid subscription, models appear in the list, but you can only use the default model.

Warning

Archived Code Assistant Pro plan does not grant access to selecting AI models.

Neurocredit consumption

The use of all models is charged in neurocredits. Different models consume different amounts of neurocredits. An indicative consumption level for each model is indicated on the AI ​​models page.

For internal models, gift neurocredits, neurocredits from the service plan, and additional paid usage are utilized. For external models, only the neurocredits included in the service plan and additional paid usage are available; you cannot use gift neurocredits for them.

Tip

You can track consumption in the Code Assistant statistics.

Use cases

  • To have the service select a model for your tasks, use the default model.
  • To select a model based on neurocredit usage, review its description and consumption rate on the AI ​​models page.
  • To ensure that organization member requests are processed within the Yandex Cloud infrastructure, enable internal models and disable external ones.

Useful links