> ## Documentation Index > Fetch the complete documentation index at: https://docs.coreweave.com/llms.txt > Use this file to discover all available pages before exploring further. # Create hot-load > Load a full or delta checkpoint into a running deployment's replicas. * The API server is `https://api.coreweave.com`. * Replace `{API_ACCESS_TOKEN}` with your [CoreWeave API access token](/security/authn-authz/manage-api-access-tokens). * For required permissions, see [IAM Access Policies](/security/iam/access-policies). The target checkpoint is named by `identity`, a single path segment beneath the deployment's configured bucket and path prefix. The bucket and prefix come from the deployment, so this request carries no storage location. A deployment runs one hot-load at a time. Creating a second while one is unfinished fails, and the same interlock blocks deployment updates while a hot-load is running. Delta checkpoints require `incrementalSnapshotMetadata`, and their resolved chain holds at most 128 deltas before a full checkpoint is needed to reset it. ```bash title="Example request" theme={{"system"}} curl -X POST https://api.coreweave.com/v1alpha1/inference/hotLoads \ -H "Content-Type: application/json" \ -H "Authorization: Bearer {API_ACCESS_TOKEN}" \ -d @data.json ``` ## OpenAPI ````yaml /openapi/inference/openapi.yaml post /v1alpha1/inference/hotLoads openapi: 3.0.3 info: title: CoreWeave Inference API version: 0.0.1 description: >- The CoreWeave Inference API provides programmatic control over inference gateways, model deployments, and capacity claims. servers: - url: https://api.coreweave.com description: CoreWeave production API. security: - bearerAuth: [] tags: - name: CapacityClaimService description: >- Endpoints for creating, listing, getting, updating, and deleting CapacityClaim reservations of GPU hardware for inference deployments. - name: DeploymentService description: >- Endpoints for creating, listing, getting, updating, and deleting model deployments. Each deployment associates a model with one or more gateways and configures runtime, resources, autoscaling, and traffic. - name: GatewayService description: >- Endpoints for creating, listing, getting, updating, and deleting inference gateways. Gateways provide authentication, request routing, load balancing, and traffic splitting for one or more deployments. - name: HotLoadService description: >- Endpoints for creating, listing, getting, and canceling checkpoint hot-loads. A hot-load loads a new set of model weights into a running deployment's replicas without restarting the inference processes or recreating the replicas. paths: /v1alpha1/inference/hotLoads: post: tags: - HotLoadService summary: Create hot-load description: >- Triggers a checkpoint hot-load on a deployment. The operation runs asynchronously, so a successful response means the request was accepted, not that the weights are loaded. operationId: HotLoadService_CreateHotLoad requestBody: required: true content: application/json: schema: $ref: '#/components/schemas/CreateHotLoadRequest' responses: '200': description: OK content: application/json: schema: $ref: '#/components/schemas/CreateHotLoadResponse' default: description: Default error response content: application/json: schema: $ref: '#/components/schemas/Status' components: schemas: CreateHotLoadRequest: description: Request for CreateHotLoad type: object properties: deploymentId: description: The ID of the deployment to hot-load type: string hotLoadId: description: >- The ID to use for the hot-load, which becomes its permanent identifier. Must be a UUID. If omitted, one is generated. Creating a second hot-load with an ID that already exists in the caller's organization fails with ALREADY_EXISTS; an ID that is unavailable for any other reason fails with INVALID_ARGUMENT. type: string identity: description: >- The identity of the target snapshot within the deployment. Must be a single path segment. Multiple hot-load operations may target the same snapshot identity. type: string type: description: >- The type of the target snapshot. `SNAPSHOT_TYPE_FULL` carries a complete set of model weights and starts a new snapshot chain. `SNAPSHOT_TYPE_INCREMENTAL` is a delta against an earlier snapshot and requires `incrementalSnapshotMetadata`. This field is required; the zero value is rejected. type: string format: enum enum: - SNAPSHOT_TYPE_FULL - SNAPSHOT_TYPE_INCREMENTAL incrementalSnapshotMetadata: description: Delta metadata, required when type is SNAPSHOT_TYPE_INCREMENTAL allOf: - $ref: '#/components/schemas/IncrementalSnapshotMetadata' promptCachePolicy: description: >- How the prompt cache is handled once the swap completes. `PROMPT_CACHE_POLICY_RESET_ALL` prevents reuse of any entry computed before the swap. `PROMPT_CACHE_POLICY_PRESERVE` allows reuse, knowingly stale under the new weights. The policy applies to requests admitted after the transition, not to requests already in flight. Omit the field rather than sending the zero value; omitting it applies `PROMPT_CACHE_POLICY_PRESERVE`. type: string format: enum enum: - PROMPT_CACHE_POLICY_RESET_ALL - PROMPT_CACHE_POLICY_PRESERVE required: - deploymentId - identity - type CreateHotLoadResponse: description: Response for CreateHotLoad type: object properties: hotLoad: description: The created hot-load allOf: - $ref: '#/components/schemas/HotLoad' readOnly: true Status: description: >- Standard error response. `code` is a [`google.rpc.Code`](https://cloud.google.com/apis/design/errors#error_codes); `message` is human-readable English; `details` carries machine-readable error details when present. type: object properties: code: description: >- The status code, which should be an enum value of [google.rpc.Code][google.rpc.Code]. type: integer format: int32 message: description: >- A developer-facing error message, which should be in English. Any user-facing error message should be localized and sent in the [google.rpc.Status.details][google.rpc.Status.details] field, or localized by the client. type: string details: description: >- A list of messages that carry the error details. There is a common set of message types for APIs to use. type: array items: $ref: '#/components/schemas/GoogleProtobufAny' IncrementalSnapshotMetadata: description: >- Metadata describing how an incremental snapshot relates to its predecessor type: object properties: previousSnapshotIdentity: description: The identity of the snapshot this delta was computed against type: string compressionFormat: description: >- The compression format of the delta. `COMPRESSION_FORMAT_ZSTD` is a Zstandard-compressed XOR delta. This field is required, and `COMPRESSION_FORMAT_ZSTD` is the only accepted value. type: string format: enum enum: - COMPRESSION_FORMAT_ZSTD checksumFormat: description: >- The checksum format used to verify snapshot integrity. This field is required, and `CHECKSUM_FORMAT_ADLER32` is the only accepted value. type: string format: enum enum: - CHECKSUM_FORMAT_ADLER32 required: - previousSnapshotIdentity - compressionFormat - checksumFormat HotLoad: description: HotLoad object with specification and status fields type: object properties: spec: description: The specification of the hot-load allOf: - $ref: '#/components/schemas/HotLoadSpec' readOnly: true status: description: The status of the hot-load allOf: - $ref: '#/components/schemas/HotLoadStatus' readOnly: true GoogleProtobufAny: description: >- Contains an arbitrary serialized message along with a @type that describes the type of the serialized message. type: object properties: '@type': description: The type of the serialized message. type: string additionalProperties: true HotLoadSpec: description: HotLoadSpec describes one immutable checkpoint hot-load operation type: object properties: id: description: The ID of the hot-load type: string readOnly: true deploymentId: description: The ID of the deployment this hot-load targets type: string readOnly: true organizationId: description: The organization ID that owns the hot-load type: string readOnly: true identity: description: >- The identity of the target snapshot within the deployment. Must be a single path segment. Multiple hot-load operations may target the same snapshot identity. type: string readOnly: true type: description: The type of the target snapshot. type: string format: enum enum: - SNAPSHOT_TYPE_UNSPECIFIED - SNAPSHOT_TYPE_FULL - SNAPSHOT_TYPE_INCREMENTAL readOnly: true incrementalSnapshotMetadata: description: Delta metadata, required when type is SNAPSHOT_TYPE_INCREMENTAL allOf: - $ref: '#/components/schemas/IncrementalSnapshotMetadata' readOnly: true promptCachePolicy: description: How the prompt cache was handled once the swap completed. type: string format: enum enum: - PROMPT_CACHE_POLICY_UNSPECIFIED - PROMPT_CACHE_POLICY_RESET_ALL - PROMPT_CACHE_POLICY_PRESERVE readOnly: true snapshotChain: description: >- The resolved snapshot chain, oldest first. The first element is always a full snapshot. type: array items: $ref: '#/components/schemas/SnapshotRef' readOnly: true HotLoadStatus: description: HotLoadStatus describes the observed state of a hot-load type: object properties: createdAt: description: The time the hot-load was created type: string format: date-time readOnly: true updatedAt: description: The time the hot-load status last changed type: string format: date-time readOnly: true finishedAt: description: The time the hot-load reached a terminal state type: string format: date-time readOnly: true state: description: >- The aggregate state of the hot-load. `HOT_LOAD_STATE_PENDING` means the operation was accepted but replica transitions have not started, and `HOT_LOAD_STATE_IN_PROGRESS` means they are underway. `HOT_LOAD_STATE_COMPLETED` means every replica serves the target snapshot identity. `HOT_LOAD_STATE_FAILED` and `HOT_LOAD_STATE_CANCELED` are terminal, and replicas may serve different snapshots until a later hot-load completes. type: string format: enum enum: - HOT_LOAD_STATE_UNSPECIFIED - HOT_LOAD_STATE_PENDING - HOT_LOAD_STATE_IN_PROGRESS - HOT_LOAD_STATE_COMPLETED - HOT_LOAD_STATE_FAILED - HOT_LOAD_STATE_CANCELED readOnly: true conditions: description: Detailed status conditions type: array items: $ref: '#/components/schemas/Condition' readOnly: true canceledAt: description: The time cancellation was requested; set only for canceled hot-loads type: string format: date-time readOnly: true SnapshotRef: description: One element of a resolved snapshot chain type: object properties: identity: description: The identity of the snapshot type: string readOnly: true type: description: The type of this snapshot in the resolved chain. type: string format: enum enum: - SNAPSHOT_TYPE_UNSPECIFIED - SNAPSHOT_TYPE_FULL - SNAPSHOT_TYPE_INCREMENTAL readOnly: true compressionFormat: description: The compression format, set for incremental snapshots. type: string format: enum enum: - COMPRESSION_FORMAT_UNSPECIFIED - COMPRESSION_FORMAT_ZSTD readOnly: true checksumFormat: description: The checksum format, set for incremental snapshots. type: string format: enum enum: - CHECKSUM_FORMAT_UNSPECIFIED - CHECKSUM_FORMAT_ADLER32 readOnly: true hotLoadId: description: >- The ID of the HotLoad operation that introduced this snapshot into the chain. type: string readOnly: true Condition: description: Condition represents a detailed status condition for resources type: object properties: type: description: Type of condition type: string readOnly: true status: description: >- The status of the condition: `STATUS_TRUE` or `STATUS_FALSE`, or `STATUS_UNSPECIFIED` when the status has not been set. type: string format: enum enum: - STATUS_UNSPECIFIED - STATUS_FALSE - STATUS_TRUE readOnly: true lastUpdateTime: description: Last time the condition was updated type: string format: date-time readOnly: true reason: description: Reason for the condition's last transition type: string readOnly: true message: description: Human-readable message indicating details about the condition type: string readOnly: true zone: description: Zone associated with the condition type: string readOnly: true securitySchemes: bearerAuth: type: http scheme: bearer bearerFormat: JWT description: CoreWeave API access token sent as a bearer token. x-default: Bearer {API_ACCESS_TOKEN} ````