> ## Documentation Index
> Fetch the complete documentation index at: https://docs.compose.market/llms.txt
> Use this file to discover all available pages before exploring further.

# Model Catalog

> Discover models, search by capability, and get pricing.

Every model on Compose is identified by a public ID in a generated catalog. The catalog carries the provider, pricing, context window, modalities, and capabilities for each model. The SDK exposes it through `sdk.models` — you discover, search, and inspect models, and the SDK routes your inference calls to the right provider automatically.

## List models

```typescript theme={null}
const { models } = await sdk.models.list();

for (const model of models) {
  console.log(model.id, model.provider, model.contextWindow);
}
```

## Search by capability

Filter by modality, provider, streaming support, context size, or price:

```typescript theme={null}
const { models } = await sdk.models.search({
  modality: "image",
  streaming: false,
  maxPrice: "0.05",
});

for (const model of models) {
  console.log(model.id, model.pricing);
}
```

## Get a single model

```typescript theme={null}
const model = await sdk.models.get("gpt-4o");

console.log(model.pricing);
console.log(model.capabilities);
```

## Get model parameters

Inspect the supported parameters for a model — temperature, top\_p, max\_tokens, and any model-specific options:

```typescript theme={null}
const params = await sdk.models.getParams("gpt-4o");

console.log(params.supported);
```

## Multi-provider routing

Models are multi-provider. The catalog may list the same upstream model under different providers, with one designated as the priority row (bare ID) and others given readable aliases:

| Public model ID | Provider | Wire model |
| --------------- | -------- | ---------- |
| `gpt-4o`        | `openai` | `gpt-4o`   |
| `azure/gpt-4o`  | `azure`  | `gpt-4o`   |

The model string you pass is the contract. The SDK resolves it to the exact provider row — it never inspects your prompt to guess a provider or silently fall back to a duplicate.

## API summary

| Method                       | Returns   | Purpose                        |
| ---------------------------- | --------- | ------------------------------ |
| `sdk.models.list()`          | `Model[]` | Browse the full catalog        |
| `sdk.models.search(filters)` | `Model[]` | Filter by capability and price |
| `sdk.models.get(id)`         | `Model`   | Inspect one model              |
| `sdk.models.getParams(id)`   | `Params`  | Supported call parameters      |

## Related

* [Inference](/sdk/inference/overview) — call any catalog model
* [Modalities](/sdk/inference/modalities) — text, image, video, speech, embeddings
* [Streaming Inference](/sdk/inference/streaming) — stream model output