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

# Smart Routing

> Set model to auto and NinjaChat picks the best model for each prompt.

## Request

```bash theme={null}
curl -X POST https://www.ninjachat.ai/api/v1/chat \
  -H "Authorization: Bearer nj_sk_YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "auto",
    "messages": [{"role": "user", "content": "Solve: x² + 5x + 6 = 0"}],
    "include_routing": true
  }'
```

## Response

```json theme={null}
{
  "model": "o3-mini",
  "choices": [{
    "message": {
      "role": "assistant",
      "content": "Factoring: x² + 5x + 6 = (x + 2)(x + 3) = 0\nSolutions: x = -2 and x = -3"
    },
    "finish_reason": "stop"
  }],
  "routing": {
    "requested": "auto",
    "resolved": "o3-mini",
    "task_type": "math",
    "reasoning": "Detected task type: math"
  },
  "cost": {"this_request": "$0.006"}
}
```

Add `"include_routing": true` to see which model was chosen and why.

## The four variants

| Model ID       | Optimizes for           | Best when...                                            |
| -------------- | ----------------------- | ------------------------------------------------------- |
| `auto`         | Quality + speed balance | You want the best model without thinking about it       |
| `auto-fast`    | Lowest latency          | Real-time apps, chatbots, low-latency pipelines         |
| `auto-cheap`   | Lowest cost             | High-volume jobs, batch processing, cost-sensitive apps |
| `auto-quality` | Highest quality         | Critical decisions, best possible output                |

## How task detection works

NinjaChat analyzes your last 3 user messages to detect the task type:

| Task type  | Detected keywords                                         | `auto` routes to    |
| ---------- | --------------------------------------------------------- | ------------------- |
| `code`     | function, debug, implement, algorithm, TypeScript, SQL... | `claude-sonnet-4.6` |
| `math`     | equation, solve, calculate, integral, probability...      | `o3-mini`           |
| `creative` | write, story, poem, imagine, fiction, lyrics...           | `gemini-3.1-pro`    |
| `analysis` | analyze, compare, evaluate, research, summarize...        | `gpt-5`             |
| `quick`    | Short prompts under 80 chars, "what is", "define"...      | `gemini-3-flash`    |
| `general`  | Everything else                                           | `gpt-5`             |

## Full routing table

<Tabs>
  <Tab title="auto (balanced)">
    | Task     | Model               |
    | -------- | ------------------- |
    | code     | `claude-sonnet-4.6` |
    | math     | `o3-mini`           |
    | creative | `gemini-3.1-pro`    |
    | analysis | `gpt-5`             |
    | quick    | `gemini-3-flash`    |
    | general  | `gpt-5`             |
  </Tab>

  <Tab title="auto-fast">
    | Task     | Model              |
    | -------- | ------------------ |
    | code     | `claude-haiku-4.5` |
    | math     | `gpt-5-mini`       |
    | creative | `gemini-3-flash`   |
    | analysis | `gpt-5-mini`       |
    | quick    | `gpt-5-mini`       |
    | general  | `gpt-5-mini`       |
  </Tab>

  <Tab title="auto-cheap">
    | Task     | Model              |
    | -------- | ------------------ |
    | code     | `deepseek-v3`      |
    | math     | `qwq-32b`          |
    | creative | `gemini-3-flash`   |
    | analysis | `deepseek-v3`      |
    | quick    | `gemini-2.5-flash` |
    | general  | `llama-4-maverick` |
  </Tab>

  <Tab title="auto-quality">
    | Task     | Model             |
    | -------- | ----------------- |
    | code     | `claude-opus-4.6` |
    | math     | `o3-mini`         |
    | creative | `gemini-3.1-pro`  |
    | analysis | `claude-opus-4.6` |
    | quick    | `claude-opus-4.6` |
    | general  | `claude-opus-4.6` |
  </Tab>
</Tabs>

## Billing

Auto variants are billed at the **resolved model's rate**. If `auto` routes to `o3-mini`, you pay $0.006. If it routes to `claude-sonnet-4.6`, you pay $0.015. The `routing` field always shows the cost-incurring model.

## Parameters

| Parameter         | Type    | Default | Description                                                          |
| ----------------- | ------- | ------- | -------------------------------------------------------------------- |
| `include_routing` | boolean | `false` | Include `routing` object in response.                                |
| `budget_cents`    | number  | —       | Override with a cost ceiling. See [Budget Routing](/budget-routing). |

## Code examples

<CodeGroup>
  ```python Python theme={null}
  import requests, os

  r = requests.post("https://www.ninjachat.ai/api/v1/chat",
      headers={"Authorization": f"Bearer {os.environ['NINJACHAT_API_KEY']}"},
      json={
          "model": "auto",
          "messages": [{"role": "user", "content": "Write a merge sort in Python"}],
          "include_routing": True,
      }
  )
  data = r.json()
  print(data["choices"][0]["choices"][0]["message"]["content"])
  print("Routed to:", data["routing"]["resolved"])  # claude-sonnet-4.6
  print("Task type:", data["routing"]["task_type"]) # code
  ```

  ```javascript Node.js theme={null}
  const r = await fetch("https://www.ninjachat.ai/api/v1/chat", {
    method: "POST",
    headers: {
      "Authorization": `Bearer ${process.env.NINJACHAT_API_KEY}`,
      "Content-Type": "application/json",
    },
    body: JSON.stringify({
      model: "auto",
      messages: [{ role: "user", content: "Prove the Pythagorean theorem" }],
      include_routing: true,
    }),
  });
  const data = await r.json();
  console.log("Routed to:", data.routing.resolved);  // o3-mini
  ```

  ```bash cURL — auto-cheap theme={null}
  curl -X POST https://www.ninjachat.ai/api/v1/chat \
    -H "Authorization: Bearer nj_sk_YOUR_API_KEY" \
    -H "Content-Type: application/json" \
    -d '{
      "model": "auto-cheap",
      "messages": [{"role": "user", "content": "Classify: positive or negative? Great product!"}],
      "include_routing": true
    }'
  ```
</CodeGroup>

## Manual model selection

If you prefer explicit control over which model runs, here's a quick reference:

```python theme={null}
models = {
    "classify": "deepseek-v3",       # $0.003 — simple tasks
    "chat":     "gpt-5",             # $0.006 — general use
    "code":     "claude-sonnet-4.6",  # $0.015 — when quality matters
    "fast":     "gemini-2.5-flash",   # $0.003 — lowest latency
}
```

See [Models](/models) for the full list with pricing and recommendations.