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

# firectl deployment create

> Reference for firectl deployment create: launch a new Fireworks on-demand deployment for a base or trained model with GPU, region, and scaling flags.

```
firectl deployment create [flags]
```

> [!WARNING]
>
> **Do not create deployments without a shape.** Always pass `--deployment-shape` — a concrete shape, or `default` to have Fireworks pick one. Deployments created without a shape skip validation, are the most common cause of failed deployment creations, and shapeless creation will soon require an explicit opt-in (`--accept-shapeless-risk`). Find shapes for your model with [`firectl deployment-shape-version match`](https://fireworks.ai/docs/tools-sdks/firectl/commands/deployment-shape-version-match), and see [Deployment shapes](https://fireworks.ai/docs/guides/ondemand-deployments.md#deployment-shapes) for details.

### Examples

```
firectl deployment create accounts/fireworks/models/falcon-7b \
  --deployment-shape default
firectl deployment-shape-version match --model accounts/fireworks/models/falcon-7b
firectl deployment create accounts/fireworks/models/falcon-7b \
  --deployment-shape <SHAPE_NAME>
firectl deployment create accounts/fireworks/models/falcon-7b \
  --accept-shapeless-risk
firectl deployment create accounts/fireworks/models/falcon-7b \
  --file=/path/to/deployment-config.json
```

Copy `<SHAPE_NAME>` from the `SHAPE NAME (version-id)` column (the resource name before the parenthesized version ID). If you use `--file`, the JSON must include `deploymentShape` set to that same `<SHAPE_NAME>` (or `"default"`).

### Flags

```
      --accelerator-count int32                  The number of accelerators to use per replica. Only use together with --deployment-shape (see warning above); setting it without a shape skips validation.
      --accelerator-type string                  The type of accelerator to use. Only use together with --deployment-shape (see warning above); setting it without a shape skips validation. Must be one of {NVIDIA_A100_80GB, NVIDIA_H100_80GB, NVIDIA_H200_141GB, NVIDIA_B200_180GB, NVIDIA_B300_288GB, AMD_MI325X_256GB, AMD_MI350X_288GB}
  -c, --cluster-id string                        The Fireworks cluster ID.
      --deployment-id string                     The ID of the deployment. If not specified, a random ID will be generated.
      --accept-shapeless-risk                    Explicit opt-out of deployment shapes: create this deployment without a pre-validated shape, which is much more likely to fail at creation. Only recommended for advanced users. Mutually exclusive with --deployment-shape.
      --deployment-shape string                  The deployment shape to use for this deployment. May be 'default' to let the server pick a default validated shape for the model.
      --deployment-template string               The deployment template to use.
      --description string                       Description of the deployment.
      --direct-route-api-keys stringArray        The API keys for the direct route. Only available to enterprise accounts.
      --direct-route-type string                 If set, this deployment will expose an endpoint that bypasses our API gateway. Must be one of {INTERNET, GCP_PRIVATE_SERVICE_CONNECT, AWS_PRIVATELINK}. Only available to enterprise accounts.
      --disable-speculative-decoding             If true, speculative decoding is disabled.
      --display-name string                      Human-readable name of the deployment. Must be fewer than 64 characters long.
      --draft-model string                       The draft model to use for speculative decoding. If the model is under your account, you can specify the model ID. If the model is under another account, you can specify the full resource name (e.g. accounts/other-account/models/falcon-7b).
      --draft-token-count int32                  The number of tokens to generate per step for speculative decoding.
      --dry-run                                  Print the request proto without running it.
      --enable-addons                            If true, enable addons for this deployment.
      --enable-session-affinity                  If true, does sticky routing based on the 'user' field. Only available to enterprise accounts.
      --file string                              Path to a JSON configuration file containing deployment settings.
  -h, --help                                     help for create
      --load-targets Map                         Map of autoscaling load metric names to their target utilization factors.
      --long-prompt                              Whether this deployment is optimized for long prompts.
      --max-context-length int32                 The maximum context length supported by the model (context window). If set to 0 or not specified, the model's default maximum context length will be used.
      --max-replica-count int32                  Maximum number of replicas for the deployment. If min-replica-count > 0 defaults to 0, otherwise defaults to 1.
      --max-with-revocable-replica-count int32   Maximum number of replicas including revocable replicas. Any replicas in excess of max_replica_count are revocable.
      --min-replica-count int32                  Minimum number of replicas for the deployment. If min-replica-count < max-replica-count the deployment will automatically scale between the two replica counts based on load.
      --ngram-speculation-length int32           The length of previous input sequence to be considered for N-gram speculation.
  -o, --output Output                            Set the output format to "text", "json", or "flag". (default text)
      --precision string                         The precision with which the model is served. If specified, must be one of {FP8, FP16, FP4, BF16}.
      --region string                            Placement: 'global', region group (us), or specific region (us-iowa-1).
      --scale-down-window duration               The duration the autoscaler will wait before scaling down a deployment after observing decreased load. Default is 10m.
      --scale-to-zero-window duration            The duration after which there are no requests that the deployment will be scaled down to zero replicas, if min-replica-count is 0. Default 1h.
      --scale-up-window duration                 The duration the autoscaler will wait before scaling up a deployment after observing increased load. Default is 30s.
      --validate-only                            If true, this will not create the deployment, but will return the deployment that would be created.
      --wait                                     Wait until the deployment is ready.
      --wait-timeout duration                    Maximum time to wait when using --wait flag. (default 1h0m0s)
```

### Global flags

```
  -a, --account-id string   The Fireworks account ID. If not specified, reads account_id from ~/.fireworks/auth.ini.
      --api-key string      An API key used to authenticate with Fireworks.
  -p, --profile string      fireworks auth and settings profile to use.
```


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