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

# Choose and switch models

> Discover models available to your llm.sg key, compare task suitability and capabilities, and switch model IDs while checking parameter compatibility.

Choose a llm.sg model based on the task you want to complete. Model IDs identify the model requested by your application; capabilities and optional parameters can differ between models.

## Discover available models

Use the platform's model directory, if provided, or request the model list:

```bash theme={null}
curl --fail-with-body --silent --show-error \
  'https://api.llm.sg/v1/models' \
  -H "Authorization: Bearer ${LLM_SG_API_KEY}"
```

Preserve the complete model ID, including prefixes, versions, and suffixes. Availability may depend on your account and key permissions. The authenticated response format still needs validation.

## Compare models

| Task | What to check |
| - | - |
| Chat and rewriting | Answer quality, response time, input and output prices |
| Coding and reasoning | Task performance, reasoning options, latency and cost |
| Long documents | Context capacity, output limits and long-input pricing |
| Image understanding | Image input support, accepted formats and limits |
| Agents and tools | Tool definitions and tool-result message support |
| Data extraction | JSON or structured output support and constraints |

Evaluate a small set of representative inputs before increasing traffic. Recheck quality, latency, and cost after changing a model.

## Switch a model

Change `LLM_SG_MODEL` in the [quickstart](https://wiki.llm.sg/en/docs/quickstart.md) to another available model ID. Begin with a minimal text request.

If your application uses images, tools, reasoning, or structured output, verify those capabilities separately. A shared endpoint does not guarantee identical parameter support.

## Router features

The platform's aggregation role does not establish support for automatic model selection, user-selected providers, failover, or configurable routing policies. Those features require product confirmation. This draft does not prescribe provider or fallback parameters.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.