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

# Manage models

> Manage your models and understand their lifecycle

## Deploying a model

To deploy a model, follow the [Deploy a model](/guides/deploy-model) guide.

## Modifying a model configuration

Select the model and open the **Configuration** tab. You can modify:

* Display name
* Hugging Face token
* Readiness level
* Scaling policy
* Worker timeout
* Pricing plan
* Pipeline-specific parameters

See [Autoscaling](/manage/autoscaling) for scaling parameter details.

## Managing a model

Select the model and open the **Manage** tab.

### Redeploying a model

Click **Redeploy** to redeploy with the current configuration. Redeploy after updating configuration, model weights, or tokenizer/processor files on the Hugging Face repository.

### Deleting a model

Click **Delete** to remove the model and all associated resources.

<Danger>
  Deleting a model is irreversible. For **Always ready** or **Super fast** deployments, all weights and prepared files are permanently removed from our infrastructure.
</Danger>

## Analytics

Open the **Analytics** tab to view insights over a selected time range:

* Number of requests
* Cost
* Instance count over time
* Tokens processed (for LLMs)
* Response time
* Pipeline-specific metrics

Use analytics to spot usage trends and tune [autoscaling](/manage/autoscaling) or [cost settings](/guides/optimizing-costs).

## Logs

Open the **Logs** tab for errors, warnings, and operational events. Check logs first when debugging deployment or inference issues — see [Troubleshooting](/support/troubleshooting).

***

## Model lifecycle

Every deployed model moves through defined lifecycle stages:

### Building

The model is being deployed on SynapsAI Cloud.

* Weights are fetched from Hugging Face.
* Quantization, optimization, or validation may run.
* Inference is not available during this stage.

### Sleeping

No instance is currently loaded.

* The model is deployed but not active.
* You can send inference requests — the platform automatically loads an instance in response.
* Load time depends on your [readiness level](/core-concepts#readiness-levels).

### Ready

At least one instance is loaded and serving inference immediately — no load delay on incoming requests.

### Failed

Deployment failed.

* This may occur due to invalid configuration, missing files, incompatible format, or internal errors.
* Resolve the issue and redeploy. If the problem persists, contact [support](/support/contacting-support).

See [Core concepts](/core-concepts#model-lifecycle) for a summary table.