Skip to content

Choosing and managing models

In Kodik settings, select the active provider and a specific model. The default auto mode lets Kodik pick the best model for each request automatically. You can also select a model manually.

Model descriptions, reasoning-effort option names and descriptions (for example, “Instant”), and per-model status notices follow the IDE display language — Russian or English; untranslated content falls back to English.

Kodik shows only the reasoning controls supported by the selected model and route. Depending on the model, this can be a Thinking/Instant toggle, an adaptive-thinking selector, or exact effort levels such as Minimal, Low, Medium, High, XHigh, or Max. Models without a documented control show no reasoning selector, and models where reasoning is mandatory do not offer Instant or Off.

The auto model has no fixed reasoning control because Kodik chooses its underlying model for each request. Gateway labels can also differ from a provider’s native vocabulary—for example, a Max level may use the provider-neutral xhigh wire value—so use the choices shown by Kodik instead of assuming every model accepts the same levels.

Mark frequently used models as favorites so they appear first in the list. This makes switching between your preferred providers faster.

If a provider exposes too many models, hide the ones you don’t need so only relevant options stay visible.

Some models consume more credits than others. Kodik shows multiplier badges so you can understand the expected cost before sending a request.

See pricing and credits for billing details.