# Introduction

> Documentation for Myles decision models and the Launchables Decision API: send a situation and typed questions, get a calibrated probability for every option in one forward pass.

Section: Get started. Web page: https://www.launchables.ai/docs

Myles models answer the small decisions inside a product or service: is this a scam, which team should take this, is this answer supported by its source. You send a **state** (the situation) and one or more **questions**. Each answer comes back with a calibrated probability for every option, read in a single forward pass, so your code decides what to automate and what to hand to a person. Myles models decide; they do not write text.

- [**Quick start**: Your first decision in one request.](https://www.launchables.ai/docs/quickstart)
- [**Question types**: noul, choice and score, with examples.](https://www.launchables.ai/docs/question-types)
- [**API reference**: Every field of POST /v1/decide.](https://www.launchables.ai/docs/api/decide)
- [**Model library**: Six models, from a browser to one GPU.](https://www.launchables.ai/docs/models)

## Two ways to call a model

- [**Myles-G: over HTTP**: POST /v1/decide with a state and typed questions. Hosted or on your own GPU.](https://www.launchables.ai/docs/model-types#gpu)
- [**CPU models: in your program**: decide(state, question, options) on your own CPU. No key, no network.](https://www.launchables.ai/docs/model-types#cpu)

## Ways to run Myles

- [**Hosted API**: Myles-G 12B and Myles-G 31B on our endpoints: £0.0015 per million input tokens, output free.](https://www.launchables.ai/docs/deploy/hosted)
- [**On your own hardware**: Servers, laptops and GPUs you operate, including disconnected networks.](https://www.launchables.ai/docs/deploy/self-hosting)
- [**In your browser**: Myles-22M on your own device, no key needed.](https://www.launchables.ai/developers)
- [**Code samples**: Python, JavaScript and curl, ready to copy.](https://www.launchables.ai/docs/samples)

## How a decision works

1. **Describe the situation.** A message, a record, a policy and a case: text or JSON.
2. **Ask typed questions.** Yes or no (`noul`), one of several labels (`choice`) or a level on a scale (`score`).
3. **Read the probabilities.** Every option gets one. Automate what clears your threshold and send the rest to a person.

Myles-G runs behind the same API on our hosted endpoint and on your own hardware, so code written against one works on the other. The CPU models take the same three things (a state, a question and options) through a function call instead; see [Model types](https://www.launchables.ai/docs/model-types).