Models and the Playground

Moth DocsApps

Models lists every model with its price against list, each model has a page of its own, Compare sets up to three side by side, and the Playground runs one and hands you the request as code.

Models, a model's page and Compare are public, so you can read every price without signing in. Trying a model, and anything about your own use, asks for your wallet.

Find a model

The Models page in Moth Console, showing the search, the four kinds with their counts, how the list stands against list prices, and a card for each model with its ID, price and saving
Models. Each card's ID is what goes in the model field of a request.
  • Type a name, an ID or a maker into Find a model. Pressing / puts the cursor there.
  • Press a kind, Chat, Image, Video or Music, to see only that kind, and press it again to see all of them. Each kind shows how many models it has and where its prices start.
  • Narrow the list by maker or by privacy mode, sort it, and switch between cards and a list.
  • The filters live in the address, so a link you share opens on the same view.

A model's card shows its maker, its ID, its Moth price, the list price struck through where its maker publishes one below the Moth price, and the saving. A model that sells above its list price says so instead. A model that is busy or paused says so on its card.

A model's own page

Pressing a card opens the model's page. Its address is console.runmoth.si/<model-id>/, so you can link straight to a model.

A model's own page in Moth Console, showing its name and ID, its price against the list price, what Moth removes before a call, its facts and the Try it card
A model's page takes the colour of its kind.
CardWhat it holds
The headThe name, one line on what the model is, the ID to copy, its privacy mode, the price against list and what Moth removes before a call
FactsThe context window or the media limits, the endpoint, the kind, the maker, the status and the date its list price was checked
Try itA message box for a chat model, with the reply, its tokens and its charge. A media model opens in the Playground, and a Sealed model in Moth Chat
Call it from your codeThe request in curl, Python and TypeScript, with Copy and a link to create a key
What a workload costsA calculator that prices your own workload per call, per day and over 30 days, with Moth and at list
Your useYour calls to this model in the last 30 days, their spend, their share of your spend and what they saved against list
ParametersWhat the model's endpoint takes for this model
The Call it from your code card on a model's page in Moth Console, showing the request with tabs for curl, Python and TypeScript and a Copy button
The request for this model, ready to copy in three languages.

Price a workload

What a workload costs starts from 2,000 tokens in and 500 out per call at 100 calls a day, and you can change each figure. For an image, video or music model the fields are images per call, seconds per clip or songs per call. The Moth price is what your balance is charged, and the list column is filled only where the maker publishes a list price.

Compare

Each card and row has a small switch that adds the model to Compare, which holds up to three. Compare lines the models up under one set of rows, with the price against list, the saving, the context window and limits, and what one sample workload costs on each, with the lowest marked. Chat samples count 2,000 tokens in and 500 out per call. The picked models travel in the address, so a comparison can be shared as a link.

The Compare page in Moth Console, showing models side by side with their prices, list prices, savings, limits and the cost of a sample workload
Compare holds up to three models.

The Playground

  1. Pick a kind and a model

    Choose Chat, Image, Video or Music, then press the model's name to pick another. Try it on a model's page opens the Playground on that model.

  2. Write the prompt and set the parameters

    A chat run takes a system line, a temperature and the most tokens the reply may use. An image run takes the number of images, a video run the clip length, and a music run a style and whether it is instrumental.

  3. Check the cost, then send

    Before you send, the page shows the most a chat run can hold, or what a media job costs. A run is charged to your balance like any API call.

  4. Read the result

    A chat reply streams in with its tokens and its exact charge. A media job shows its progress and then its files, and nothing is saved, so download what you want to keep.

The Playground in Moth Console, showing the kind and the model, a prompt that was sent with its reply, the parameters and what the run cost
The Playground after a run, with the reply and what it cost.

Under the prompt, a line shows what the model receives, which is your prompt and nothing about you. The request as code follows every change in curl, Python and TypeScript, so what you copy sends exactly what the box holds.

Sessions stay in the browser. Each chat run is saved as a session in this browser only, with its reply, tokens and charge. Moth keeps no prompts or replies, so clearing the browser's data clears the sessions, and a deleted session can't be brought back.