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

# Swytcho AI Models: Capabilities and Selection

> Explore the AI models available on Swytcho, understand their capabilities and trade-offs, and choose the right model for your use case and budget.

Swytcho gives you access to a curated set of AI models through a single, OpenAI-compatible API. Each model family is optimized for a different balance of capability, speed, and cost — so you can pick the right tool for every task without changing your integration code.

## Model Families

<CardGroup cols={3}>
  <Card title="swytcho-1" icon="brain">
    **General purpose** — Swytcho's flagship model. Best for complex reasoning, long-context tasks, nuanced instruction following, and production workloads where quality matters most.
  </Card>

  <Card title="swytcho-1-mini" icon="bolt">
    **Fast & cost-efficient** — A smaller, distilled model optimized for low latency and high throughput. Ideal for classification, summarization, and high-volume pipelines.
  </Card>

  <Card title="swytcho-embed" icon="vector-square">
    **Embeddings** — Converts text into high-dimensional vector representations. Purpose-built for semantic search, retrieval-augmented generation (RAG), and similarity scoring.
  </Card>
</CardGroup>

### At a Glance

| Model            | Context Window | Best For                            | Relative Cost |
| ---------------- | -------------- | ----------------------------------- | ------------- |
| `swytcho-1`      | 128 000 tokens | Complex reasoning, long documents   | \$\$\$        |
| `swytcho-1-mini` | 32 000 tokens  | Summarization, classification, chat | \$            |
| `swytcho-embed`  | 8 192 tokens   | Embeddings, semantic search, RAG    | \$            |

## List Available Models

To fetch the full list of models currently available on your account, send a `GET` request to the `/v1/models` endpoint. The response follows the standard OpenAI models list format.

```bash theme={null}
curl https://api.swytcho.com/v1/models \
  -H "Authorization: Bearer $SWYTCHO_API_KEY"
```

A successful response returns an array of model objects:

```json theme={null}
{
  "object": "list",
  "data": [
    {
      "id": "swytcho-1",
      "object": "model",
      "created": 1713000000,
      "owned_by": "swytcho"
    },
    {
      "id": "swytcho-1-mini",
      "object": "model",
      "created": 1713000000,
      "owned_by": "swytcho"
    },
    {
      "id": "swytcho-embed",
      "object": "model",
      "created": 1713000000,
      "owned_by": "swytcho"
    }
  ]
}
```

<Note>
  Always query the `/v1/models` endpoint to discover the latest available models. New versions and experimental previews are added here first, before they are reflected in this documentation.
</Note>

## Choosing the Right Model

Use the decision guide below to match your task to the best model.

<Steps>
  <Step title="Identify your task type">
    * **Open-ended generation, reasoning, or coding** → start with `swytcho-1`
    * **High-volume classification, extraction, or chat** → start with `swytcho-1-mini`
    * **Semantic search, RAG retrieval, or similarity** → use `swytcho-embed`
  </Step>

  <Step title="Estimate your context length">
    If your combined prompt and expected output exceed 32 000 tokens, you need `swytcho-1`. For shorter exchanges, `swytcho-1-mini` is faster and cheaper.
  </Step>

  <Step title="Prototype with swytcho-1, optimize with swytcho-1-mini">
    Build your feature with `swytcho-1` first to establish a quality baseline. Once you are satisfied with the results, test `swytcho-1-mini` against the same inputs — many tasks perform equally well at a fraction of the cost.
  </Step>

  <Step title="Benchmark before committing">
    Run both models against a representative sample of your real data. Measure accuracy, latency, and cost per request, then choose the model that meets your thresholds.
  </Step>
</Steps>

## Model Versioning and Stability

Swytcho uses a **date-pinned alias** system to give you control over model updates:

* **Floating aliases** (e.g., `swytcho-1`) always resolve to the latest stable release. Use these during development for access to the newest improvements.
* **Date-pinned versions** (e.g., `swytcho-1-2025-04-01`) are frozen snapshots. Use these in production when you need deterministic behavior across deployments.

When Swytcho retires a model version, it sends deprecation notices at least **90 days in advance** via email and dashboard banner. The floating alias is automatically migrated to the next stable release on the deprecation date.

<Tip>
  Pin to a date-stamped model version in your production configuration and update it deliberately during planned maintenance windows. This prevents unexpected behavior changes from silent alias rotations.
</Tip>