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

# Models

> The catalog is deliberately curated: production models spanning reasoning, coding, chat, and vision, all behind the same OpenAI-compatible endpoint.

Larger models give you more capability on complex tasks; smaller ones respond faster and cost less per token. Capability is never one-to-one across model families, so treat these recommendations as starting points and spot-check candidates on your own prompts, the [playground](https://platform.pearlresearch.ai/playground) makes side-by-side comparisons quick.

## Your first request

Every model answers on the same OpenAI-compatible endpoint, so moving between them is a one-word change. This example uses `deepseek-ai/DeepSeek-V4-Pro-0813` because it is the first entry in the live catalog below, not because it is the best fit for your workload; any other `id` from the table drops straight in.

```bash theme={null}
curl https://inference.pearlresearch.ai/v1/chat/completions \
  -H "Authorization: Bearer $PEARL_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-ai/DeepSeek-V4-Pro-0813",
    "messages": [{"role": "user", "content": "Hello"}]
  }'
```

## Supported models

Pass the model ID as the `model` parameter in your requests. Context is shown as rounded thousands; the exact value, and each model's max output, are at `GET /models`. Quantization is the serving precision the deployment runs at, and Reasoning lists the effort levels the model serves plus the default you get by omitting the control. A level the model doesn't list rounds up to the next one it does, or to its highest above the top (see [Reasoning models](/docs/reasoning.md)).

<table>
  <thead>
    <tr>
      <th>Model</th>
      <th>Context</th>
      <th>Quantization</th>
      <th>Input</th>
      <th>Features</th>
      <th>Reasoning</th>
    </tr>
  </thead>

  <tbody>
    <tr>
      <td>DeepSeek V4 Pro 0813\
      `deepseek-ai/DeepSeek-V4-Pro-0813`</td>
      <td>1M</td>
      <td>FP8</td>
      <td>Text</td>
      <td>JSON mode, Reasoning, Structured outputs, Tool calling</td>
      <td>high, max (default high)</td>
    </tr>

    <tr>
      <td>DeepSeek V4.1 Flash\
      `deepseek-ai/DeepSeek-V4.1-Flash`</td>
      <td>1M</td>
      <td>FP8</td>
      <td>Text, Image</td>
      <td>JSON mode, Reasoning, Structured outputs, Tool calling</td>
      <td>low, high, xhigh, max (default high)</td>
    </tr>

    <tr>
      <td>DeepSeek V4.1 Flash 1017\
      `deepseek-ai/DeepSeek-V4.1-Flash-1017`</td>
      <td>1M</td>
      <td>FP8</td>
      <td>Text, Image</td>
      <td>JSON mode, Reasoning, Structured outputs, Tool calling</td>
      <td>low, high, xhigh, max (default high)</td>
    </tr>

    <tr>
      <td>Qwen3.8-27B\
      `Qwen/Qwen3.8-27B`</td>
      <td>256k</td>
      <td>FP8</td>
      <td>Text, Image</td>
      <td>Reasoning, Structured outputs, Tool calling</td>
      <td>low, medium, xhigh (default xhigh)</td>
    </tr>

    <tr>
      <td>GLM-5.3\
      `zai-org/GLM-5.3`</td>
      <td>1M</td>
      <td>FP4</td>
      <td>Text</td>
      <td>JSON mode, Reasoning, Structured outputs, Tool calling</td>
      <td>low, high, max (default max, always on)</td>
    </tr>

    <tr>
      <td>GLM-5.3-Flash\
      `zai-org/GLM-5.3-Flash`</td>
      <td>1M</td>
      <td>FP4</td>
      <td>Text, Image</td>
      <td>JSON mode, Reasoning, Structured outputs, Tool calling</td>
      <td>low, high, max (default max, always on)</td>
    </tr>
  </tbody>
</table>

> [!NOTE]
>
> Pricing is per token and varies by model. Live rates for your organization, including discounted cached-input pricing, are always on the dashboard's [Models](https://platform.pearlresearch.ai/models) page.

## List models via the API

The live catalog is also served by the standard OpenAI models endpoint. Each entry's `id` is the value you pass as `model`, alongside per-model context length, modalities, and supported parameters:

```bash theme={null}
curl https://inference.pearlresearch.ai/v1/models \
  -H "Authorization: Bearer $PEARL_API_KEY"
```

## Migrating from closed models

Moving from a closed-model provider? Because the endpoint is OpenAI-compatible, migrating is usually a base-URL and model-name change. As a starting point:

| If you use | Start with |
| - | - |
| Frontier coding & agent models (GPT-5-class, Claude Opus-class) | DeepSeek V4 Pro |
| Mid-tier general models (GPT-4o-class, Claude Sonnet-class) | GLM-5.3 |
| Fast, low-cost models (mini/Haiku-class) | GLM-5.3 Flash |
| Fast multimodal models (GPT-4o-mini-class) | GLM-5.3 Flash |
| Multimodal models used for image understanding | GLM-5.3 Flash |


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