Browse Models
Visual Model Catalog — Search, filter, and compare every model with real-time pricing, context windows, and capabilities.
Using a Model
To use any model, pass its ID to the chat completions endpoint:Programmatic Access
Get the complete model list via API:Meta Muse Spark Contributor
meta/muse-spark-1.3-contributor is an opt-in, direct Meta model. It is not
zero-data-retention: Meta may use submitted prompts, files, context, and model
responses to train its models. Applications must not infer permission to send
customer content merely from the absence of a BAA.
The model record explicitly includes zdr: false,
compliance: { zdr: false, baa: false, hipaa: false }, and
data_policy: { training: true, zero_data_retention: false }.
Clients must inspect these fields rather than assuming every catalog model is ZDR.
Case.dev prices per million tokens are $0.20 input**, **$0.004 cached input,
and $0.40 output. Each is twice Meta’s provider rate; cached tokens are
subtracted from ordinary input before billing. The same customer charge is
returned in usage.cost and recorded in billable usage for both streaming and
non-streaming requests. Downstream applications need no additional multiplier.
The route rejects policies requiring ZDR, BAA, or HIPAA, honors provider
restrictions, and fails closed when policy lookup is unavailable. The model is
listed only where META_AI_API_KEY is configured on the router. It is never a
automatic fallback for another model. Existing models and prices are unchanged.
Provider references: Meta model documentation
and pricing and Contributor data use.
