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

Get pricing

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

Upstream list prices for every certified route, and the platform fee applied on top. The same data is on Pricing.

Response

  • markup_basis_pointsintegerrequired

    Platform fee on upstream cost for your team, in basis points. 750 is 7.5%.

  • models[]arrayrequired

    One entry per certified route: a model and a provider that serves it, so a model with three providers appears three times. Chat models first, then evaluation, embedding and image models.

  • models[].input_per_tokenstringrequired

    Upstream USD per input token.

  • models[].output_per_tokenstringrequired

    Upstream USD per output token. 0 for embedding and image models.

  • models[].cache_read_per_token, cache_write_per_tokenstringoptional

    Upstream USD per cached input token read, and per token written to the cache, on routes that price caching.

  • models[].web_search_per_callstringoptional

    Upstream USD per web search, on routes that search. See Web search.

  • models[].per_imagestringoptional

    Image models: upstream USD per image at the provider's default size. See Image generation.

  • models[].per_image_tiersarrayoptional

    Image models that price by size, quality or style: one entry per listed price, each with cost and any of size, quality and style.

  • models[].image_output_tokensarrayoptional

    Token-priced image models (OpenAI's gpt-image, MAI): the output tokens one image took when measured, per size (auto when none is sent) and quality, billed at output_per_token. Only these sizes are accepted. per_image_extra, when present, is held on every image for an operation a provider option can turn on.

  • models[].notestringrequired

    What the prices apply to.

Prices are the provider's list prices; the charge for a request is settled from what the provider billed for it, plus the fee. See What you pay.