Models

This content is not available in your language yet.

Call GET /open/v1/models with a bearer token that grants chat.play. The response describes the models available to the conversation API, with public identifiers and usage rates. Use these identifiers in requests rather than upstream provider names.

The initial model is deepseek-v4-flash. Input, cached input and output have separate rates, expressed in credits per million tokens. Catalog estimates are bounds, not a fixed charge per message; the final charge is recorded on the completed operation.

Select a model

Read GET /open/v1/player/role-settings?roleId=…, then save selectModel through POST /open/v1/player/role-settings/save with the same roleId. Use only a model returned by the catalog.

Use the returned variants and capability fields as the authority for currently available routes. Do not infer automatic cross-lane failover from a model being listed.

For Agent mode, supportsMultiPass indicates tool support. Read /open/v1/player/agent-mode for runtime availability and the selected model’s eligibility. Free Seed/basebot models and routes known not to support tools are excluded. A multi-call Agent execution can exceed the catalog’s per-turn estimate; see usage and billing.

See Conversation loop to make a request and Quotas and pricing to read its cost.