# Managed Models

Source: https://docs.sliplane.io/inference/managed-models

Use a managed model to always get the best current model for your use case.

> **info**
>
> **Private beta:** Email [support@sliplane.io](mailto\:support@sliplane.io) to request Inference access for your team.

A managed model is a model ID that Sliplane keeps pointed at the best current model for its purpose. Request `sliplane/frontier`, `sliplane/balanced`, or `sliplane/cheap` instead of a specific model. When better models are released, Sliplane switches the managed model to the new one. Your code stays the same.

| Managed model       | Use it for                                                        |
| ------------------- | ----------------------------------------------------------------- |
| `sliplane/frontier` | The most capable model available, for complex reasoning and code  |
| `sliplane/balanced` | Strong quality at a moderate price, for most production workloads |
| `sliplane/cheap`    | The lowest price, for high-volume and simple tasks                |

Use managed models like any other model, with `/v1/chat/completions` and `/v1/responses`:

```sh
curl https://inference.sliplane.io/v1/chat/completions \
  -H "Authorization: Bearer $SLIPLANE_INFERENCE_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "sliplane/frontier",
    "messages": [{ "role": "user", "content": "Hello!" }]
  }'
```

## Current models

Each managed model resolves to a model of your key's [region](/inference/guardrails#region). The same ID works in every region.

| Managed model       | Global             | EU              | Confidential |
| ------------------- | ------------------ | --------------- | ------------ |
| `sliplane/frontier` | `claude-fable-5-1` | `kimi-k3.1-pro` | `glm-5.3`    |
| `sliplane/balanced` | `glm-5.3`          | `glm-5.3`       | `glm-5.3`    |
| `sliplane/cheap`    | `claude-haiku-5-5` | `glm-5.3-flash` | `gemma4-26b` |

Every model behind a managed model supports streaming and function calling.

## Pricing

A request to a managed model costs the same as a request to the model it resolves to. See [Models & Pricing](/inference/models) for the current rates.

## Model changes

Sliplane announces every change by email to team owners at least seven days before it takes effect. The announcement names the new model and its prices. After the change, requests to the managed model use the new model and its prices.

To keep using a model after a change, request that model by its own ID.

## Responses and usage

The `model` field of every response contains the model that handled the request, such as `claude-fable-5-1`. [Usage](/inference/usage) lists each request under that model.

## Guardrails

A [guardrail](/inference/guardrails#models) can allow managed models like any other model. Allowing a managed model allows the model it currently resolves to. `GET /v1/models` lists the managed models of the key's region.