API reference
List models
Dashboard
BetaWe just launched Apps: hosted models, actions and pages, on the same bill as your inference.Read the Apps docs

List models

GEThttps://api.inferencebureau.com/v1/models

The certified models this key may call after every Directive is applied, one entry per model: language models first, then evaluation models, embedding models and image models, then your team's and project's aliases. With allowedModels, deniedModels, requireZdr or requireNoTraining set, models your key can't use are left out, and so are aliases that name them. A capability is listed when at least one provider your key may use supports it, and requests that need it are routed only to those providers. The public catalog is on Models.

Model object

  • idstringrequired

    Slug to pass as model.

  • typestringrequired

    language for chat models, evaluation for models called with POST /v1/evaluate, embedding for models called with POST /v1/embeddings, image for models called with POST /v1/images/generations.

  • owned_bystringrequired

    The model's creator; org for your team's aliases.

  • alias_forstringoptional

    Aliases only: the model the alias names. Every other field is that model's.

  • routing_profilestringoptional

    Aliases that choose their model through a routing profile only: the profile's id. alias_for is then the profile's baseline, and each request may run on another model of its pool.

  • context_windowintegeroptional

    Language models: maximum input plus output tokens, on the largest provider you may use.

  • max_output_tokensintegeroptional

    Language models: maximum generated tokens. Requests without max_tokens are limited to 4,096.

  • capabilitiesobjectoptional

    Language models: booleans for tools, structuredOutput, streaming, and vision (image input). webSearch is "tool", "built_in", or false, and webSearchTools lists the web_search tool versions the model accepts; see Web search.

  • endpointstringoptional

    Evaluation, embedding and image models: the endpoint that serves them.

  • editsbooleanoptional

    Image models: whether they also take POST /v1/images/edits.

  • zero_data_retentionbooleanoptional

    Evaluation models: whether a provider you may use offers zero data retention.