Prices

Which model

Three models for what you are making

Five short questions. Every answer is a filter on what the catalog publishes and what this gateway measured. Nothing here is a review.

What is it for

How long are the inputs

Does it read pictures or files

Does it call tools or return JSON

What may a message cost

A typical message is 500 tokens sent and 500 written, at the billed price.

111 models fit all five

  1. Cheapest that fits

    Mistral Nemo

    mistralai/mistral-nemo

    The lowest price for a typical message among every model that fits: less than 1/100 of a cent.

    A typical message
    less than 1/100 of a cent · $0.0319 in, $0.033 out per 1Mbilled price
    Context window
    131k tokensthe catalog's word
    Tool calls
    Declaredthe provider's word
    Try it in chatGet a key

  2. From a lab you know

    GPT-6 Luna Pro

    openai/gpt-6-luna-pro

    OpenAI's newest model that fits, from the labs most people know.

    A typical message
    about 0.03¢ · $0.11 in, $0.55 out per 1Mbilled price
    Context window
    1M tokensthe catalog's word
    Tool calls
    Declaredthe provider's word
    Try it in chatGet a key

How the three are chosen

The first is the cheapest model that meets every choice. The second is the one this gateway measured doing the task, from its own code, extraction and reasoning checks, or, for chat and writing, the fastest first token it measured. The third is the newest model from the best known lab that fits. Free models are left out: their limits are not ours to promise.

Every price here is the upstream's published price plus 10%, the same for every account.

Every model and its price, or the cheapest model for one need.

Antimatter