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

# TrueFoundry Managed Models

> Add curated open-weight and frontier models to TrueFoundry's AI Gateway without a provider account or API key.

TrueFoundry Managed Models are served on TrueFoundry's own upstream capacity. Unlike other
providers, there is no account to create with the model vendor and no API key to configure or
rotate. Usage is billed to a prepaid credit balance.

### Adding Models

<Steps>
  <Step title="Enable TrueFoundry Managed Models">
    From the TrueFoundry dashboard, navigate to **AI Gateway** > **Models** and click
    **Start with free credit**.

    <Frame caption="TrueFoundry Managed Models on the Models page">
      <img src="https://mintcdn.com/truefoundry/tyxjlHUoO4j_JVJH/images/truefoundry-managed-get-started.png?fit=max&auto=format&n=tyxjlHUoO4j_JVJH&q=85&s=1025edb5fa2795524997e99fcceb8703" alt="Models page with the TrueFoundry Managed Models card" width="3016" height="1734" data-path="images/truefoundry-managed-get-started.png" />
    </Frame>

    No card is required to start.
  </Step>

  <Step title="Select Models">
    Select the models you want to enable. Input and output pricing per million tokens and the
    context window are shown for each model. Click **Submit and Enable** to finish.

    <Note>
      There are no credentials to enter. You can change the selection at any time.
    </Note>

    <Frame caption="Models Selection">
      <img src="https://mintcdn.com/truefoundry/tyxjlHUoO4j_JVJH/images/truefoundry-managed-select.png?fit=max&auto=format&n=tyxjlHUoO4j_JVJH&q=85&s=de4c64d00e523ac9f7f5a647e20552ff" alt="Model selection with input, output pricing and context window" width="3016" height="1734" data-path="images/truefoundry-managed-select.png" />
    </Frame>
  </Step>

  <Step title="Set Access Control">
    Configure who can access these models from **Access Control**. Learn more about access control
    [here](/docs/ai-gateway/gateway-access-control).
  </Step>
</Steps>

To add or remove models later, go to **Manage Accounts** > **TrueFoundry Managed** and click
**Add model**.

<Frame caption="Connected Providers & Accounts">
  <img src="https://mintcdn.com/truefoundry/tyxjlHUoO4j_JVJH/images/truefoundry-managed-accounts.png?fit=max&auto=format&n=tyxjlHUoO4j_JVJH&q=85&s=9597deac084fe4e491a83a861ed53188" alt="TrueFoundry Managed account with its enabled models" width="3016" height="1734" data-path="images/truefoundry-managed-accounts.png" />
</Frame>

### Inference

After adding the models, you can perform inference using an OpenAI-compatible API via the Playground
or by integrating with your own application.

```python Python theme={"dark"}
from openai import OpenAI

client = OpenAI(
    api_key="your-truefoundry-api-key",
    base_url="{GATEWAY_BASE_URL}",
)

response = client.chat.completions.create(
    model="tfy-managed/grok-4.7",
    messages=[{"role": "user", "content": "Summarise this incident report."}],
)
```

Each model's **How To Use** tab gives you the Base URL, Model ID and API Key, with ready-made
snippets for the OpenAI library, LangChain, Node.js, curl and others, in both streaming and
non-streaming form.

<Note>
  Model identifiers are stable. Retirement of a model is announced in the changelog with a named
  replacement and a migration window.
</Note>

### Supported APIs

Once your TrueFoundry Managed Models account is configured, the following API surfaces are
available through the AI Gateway.

| API | Endpoint | Tracing | Cost Tracking |
| - | - | - | - |
| Chat Completions | `/chat/completions` | **✅** | **✅** |
| Embeddings | `/embeddings` | **✅** | **✅** |

### Credits

Managed model usage draws down a prepaid credit balance. Credit is bought as a separate payment and
does not appear on your plan invoice. Manage it at **Settings** > **TrueFoundry Model Credits**.

<Frame caption="TrueFoundry Model Credits">
  <img src="https://mintcdn.com/truefoundry/tyxjlHUoO4j_JVJH/images/truefoundry-managed-credits.png?fit=max&auto=format&n=tyxjlHUoO4j_JVJH&q=85&s=1c62cf705bfbf95cfc942aaec31af82c" alt="Credit balance, add credits, and automatic top-up" width="3016" height="1734" data-path="images/truefoundry-managed-credits.png" />
</Frame>

<Steps>
  <Step title="Add credits">
    Select a preset amount or enter your own, then click **Add Credits**. Credit does not expire.
  </Step>

  <Step title="Enable automatic top-up">
    Set a balance threshold and a top-up amount, then click **Save Changes**. Credit is added
    automatically from your saved card whenever the balance falls below the threshold.

    <Note>
      Automatic top-up requires a saved card. A declined charge is retried and you are notified by
      email. After repeated failures, automatic top-up is disabled so your card is not retried
      indefinitely.
    </Note>
  </Step>
</Steps>

Every purchase appears under **TrueFoundry Managed Credit History** with its date, invoice number,
amount, status, and whether it was triggered by a user or by automatic top-up. Receipts are
available per transaction.

<Warning>
  At a zero balance, managed models stop serving requests and return an error. Other provider
  accounts in your gateway are unaffected. Adding credit restores service immediately, and your
  model selection is unchanged. You are notified by email before and when this happens.
</Warning>

<Info>
  Managed model usage can be billed against your existing TrueFoundry invoice instead of a card.
  Contact your account team to enable invoiced billing.
</Info>

### Data handling

No provider used by TrueFoundry Managed Models trains on your prompts or completions. This applies
to every model in the catalog.

Zero data retention (ZDR) means prompts and completions are never written to storage after a
request is served.

| Retention | Models |
| - | - |
| **Zero data retention** | `deepseek-v4-flash-0731`, `deepseek-v4-pro-0813`, `deepseek-v4.1-flash`, `glm-5.3`, `glm-5.3-flash`, `gpt-oss-120b`, `kimi-k3`, `minimax-m3`, `nemotron-3-ultra`, `qwen3.8-max` |
| **Up to 30 days** | All other models. Retained for abuse monitoring only, and never used for training. |

<Note>
  Inference is currently performed in the United States. Region-specific routing is on the roadmap.
</Note>

### Reliability

Each model is served by multiple upstream providers with automatic failover. If a provider becomes
unavailable or degrades, requests are routed to the next provider with no change on your side.

Quantization is fixed per model, so a model identifier always returns the same precision regardless
of which provider serves the request.


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