> ## Documentation Index > Fetch the complete documentation index at: https://docs.costgraph.ai/llms.txt > Use this file to discover all available pages before exploring further. # List accelerators > Returns items and a total for the accelerators held by the tenant selected by the X-CostGraph-Tenant-ID header. An accelerator that has reported gets its own item carrying its utilisation band, its 24h, 7d and 30d utilisation and memory, its hourly and monthly cost where the catalog prices it, and its MIG partitions with the ones that ran no compute marked idle. The accelerators on a machine that have not reported are counted in one further item per machine, with a null gpu_uuid, unreported_card_count set, and every measured field null rather than zero. A machine that reports its accelerators through the host agent rather than through Kubernetes gets its own item in the same shape, counted in unreported_card_count with every measured field null, since the agent reports what the hardware is and never how busy it was; such a machine is listed separately from any Kubernetes node and is never merged with one. model_source names what answered for the model, and reported_by names which path answered for the accelerator: operator for a card reporting through Kubernetes, agent for one the host agent inventoried, and catalog for cards the machine is priced for but none has reported. placement says where the accelerator sits and carries the ids a caller links on: placement.instance_id always, and placement.cluster_id, placement.node_id and placement.node_name where the machine is a Kubernetes node we hold, which is also what makes placement.kind workload rather than instance. placement.namespace names the namespace of the workload that asked for the accelerator, and stays null where no workload on that node asks for one or where more than one namespace does. gpu_availability states every signal the card could report, read from the measurements stored against it: supported where it was measured, capacity_unknown where it was measured but nothing expresses it as a percentage, unsupported where the card reports other signals but not this one, and no_data where the card reported nothing at all; no signal is ever omitted or returned as zero. A report whose 7-day or 30-day window has not been computed carries window_7d or window_30d as null rather than as a zero measurement. cost carries the currency every amount is in, the hourly and projected monthly rate the catalog prices the card at, and mtd_reason saying why mtd and change_percent are null; spend to date is billed for the whole machine, so it is never attributed to one accelerator and never rendered as zero. utilization_sparkline and memory_utilization_sparkline each carry the trailing 24 hours as 12 fixed-width 2 hour buckets, oldest first, each holding the highest percentage the card reached in that bucket; a bucket the card took no sample in is null, a card that has never reported carries twelve nulls rather than zeros, and a card carved into MIG partitions is read once for the whole card rather than once per partition. total_count is the filtered count before paging. ## OpenAPI ````yaml /api-reference/costgraph/openapi.json get /api/v1/tenant/gpus openapi: 3.0.0 info: description: Read and manage your CostGraph organization, spend, alerts, and settings. title: CostGraph API contact: {} version: '1.0' servers: - url: https://api.costgraph.ai security: [] tags: - name: ai x-group: AI - name: ai-serving x-group: AI serving - name: anomalies x-group: Anomalies - name: auth x-group: Auth - name: billing x-group: Billing - name: billing-export x-group: Billing export - name: budgets x-group: Budgets - name: ci x-group: CI - name: compute-recommendations x-group: Compute recommendations - name: config x-group: Config - name: cost x-group: Cost - name: gpus x-group: GPUs - name: graphai x-group: Graph AI - name: infracost x-group: Infracost - name: integrations x-group: Integrations - name: invitations x-group: Invitations - name: kubernetes-clusters x-group: Kubernetes clusters - name: marketplace x-group: Marketplace - name: network-requests x-group: Network requests - name: notifications x-group: Notifications - name: oauth x-group: OAuth - name: oauth-clients x-group: OAuth clients - name: opencost x-group: OpenCost - name: organization x-group: Audit log - name: organizations x-group: Organizations - name: placement-alternatives x-group: Placement alternatives - name: reports x-group: Reports - name: service-map x-group: Service map - name: settings x-group: Settings - name: sso x-group: Single sign-on - name: tenants x-group: Tenants - name: user x-group: Users - name: virtual-machines x-group: Virtual machines - name: virtual-tags x-group: Virtual tags - name: workflows x-group: Workflows paths: /api/v1/tenant/gpus: get: tags: - gpus summary: List accelerators description: >- Returns items and a total for the accelerators held by the tenant selected by the X-CostGraph-Tenant-ID header. An accelerator that has reported gets its own item carrying its utilisation band, its 24h, 7d and 30d utilisation and memory, its hourly and monthly cost where the catalog prices it, and its MIG partitions with the ones that ran no compute marked idle. The accelerators on a machine that have not reported are counted in one further item per machine, with a null gpu_uuid, unreported_card_count set, and every measured field null rather than zero. A machine that reports its accelerators through the host agent rather than through Kubernetes gets its own item in the same shape, counted in unreported_card_count with every measured field null, since the agent reports what the hardware is and never how busy it was; such a machine is listed separately from any Kubernetes node and is never merged with one. model_source names what answered for the model, and reported_by names which path answered for the accelerator: operator for a card reporting through Kubernetes, agent for one the host agent inventoried, and catalog for cards the machine is priced for but none has reported. placement says where the accelerator sits and carries the ids a caller links on: placement.instance_id always, and placement.cluster_id, placement.node_id and placement.node_name where the machine is a Kubernetes node we hold, which is also what makes placement.kind workload rather than instance. placement.namespace names the namespace of the workload that asked for the accelerator, and stays null where no workload on that node asks for one or where more than one namespace does. gpu_availability states every signal the card could report, read from the measurements stored against it: supported where it was measured, capacity_unknown where it was measured but nothing expresses it as a percentage, unsupported where the card reports other signals but not this one, and no_data where the card reported nothing at all; no signal is ever omitted or returned as zero. A report whose 7-day or 30-day window has not been computed carries window_7d or window_30d as null rather than as a zero measurement. cost carries the currency every amount is in, the hourly and projected monthly rate the catalog prices the card at, and mtd_reason saying why mtd and change_percent are null; spend to date is billed for the whole machine, so it is never attributed to one accelerator and never rendered as zero. utilization_sparkline and memory_utilization_sparkline each carry the trailing 24 hours as 12 fixed-width 2 hour buckets, oldest first, each holding the highest percentage the card reached in that bucket; a bucket the card took no sample in is null, a card that has never reported carries twelve nulls rather than zeros, and a card carved into MIG partitions is read once for the whole card rather than once per partition. total_count is the filtered count before paging. parameters: - description: Tenant ID name: X-CostGraph-Tenant-ID in: header required: true schema: type: string - description: Filter to one accelerator model name: model in: query schema: type: string - description: Filter to one utilisation band name: utilization_band in: query schema: type: string - description: Filter to accelerators held by one instance name: instance_id in: query schema: type: string - description: Filter to accelerators on the machines of one Kubernetes cluster name: cluster_id in: query schema: type: string - description: Filter to one or more cloud providers name: provider in: query explode: true schema: type: array items: type: string - description: 'Filter to where the accelerator sits: instance or workload' name: placement_kind in: query explode: true schema: type: array items: type: string - description: >- Case-insensitive match over the accelerator model and the name of the machine holding it name: search in: query schema: type: string - description: Filter to partitioned or whole cards name: partitioned in: query schema: type: boolean - description: >- Only accelerators that reported within this window, as a duration such as 24h, 7d or 30d. Omit to include ones that stopped reporting. name: window in: query schema: type: string - description: Number of rows to return (default 50, max 200) name: limit in: query schema: type: integer - description: Rows to skip name: offset in: query schema: type: integer responses: '200': description: OK content: application/json: schema: allOf: - $ref: '#/components/schemas/responses.SuccessResponse' - type: object properties: data: $ref: '#/components/schemas/db.GPUListPage' '400': description: Bad Request content: application/json: schema: $ref: '#/components/schemas/responses.ErrorResponse' '401': description: Unauthorized content: application/json: schema: $ref: '#/components/schemas/responses.ErrorResponse' '403': description: Forbidden content: application/json: schema: $ref: '#/components/schemas/responses.ErrorResponse' '500': description: Internal Server Error content: application/json: schema: $ref: '#/components/schemas/responses.ErrorResponse' security: - BearerAuth: [] components: schemas: responses.SuccessResponse: type: object required: - message - status properties: data: {} message: type: string example: some message status: type: string example: success db.GPUListPage: type: object required: - items - limit - offset - total_count properties: items: type: array items: $ref: '#/components/schemas/db.GPUListRow' limit: type: integer offset: type: integer total_count: type: integer responses.ErrorResponse: type: object required: - message - status properties: message: type: string example: some message status: type: string example: error db.GPUListRow: type: object required: - memory_utilization_sparkline - utilization_sparkline properties: action: type: string cost: $ref: '#/components/schemas/db.GPUCost' cost_per_hour: type: number data_age_hours: type: number data_quality_state: type: string gpu_count: type: number gpu_index: type: integer gpu_uuid: type: string idle_partition_count: type: integer instance_id: type: string last_reported_at: type: string memory_used_p95: type: number memory_used_p95_30d: type: number memory_used_p95_7d: type: number memory_utilization_percent: type: number memory_utilization_sparkline: type: array items: type: number mig_enabled: type: boolean model: type: string model_source: type: string partition_count: type: integer partitions: type: array items: $ref: '#/components/schemas/db.GPUPartition' placement: $ref: '#/components/schemas/db.GPUPlacement' power_mean_watts: type: number provider: type: string region: type: string reported_by: type: string unreported_card_count: type: integer urgency: type: string utilization_band: type: string utilization_p95_30d: type: number utilization_p95_7d: type: number utilization_percent: type: number utilization_sparkline: type: array items: type: number vendor: type: string db.GPUCost: type: object properties: change_percent: type: number currency: type: string hourly_rate: type: number mtd: type: number mtd_reason: type: string projected_monthly_spend: type: number db.GPUPartition: type: object properties: idle: type: boolean mig_instance_id: type: string mig_profile: type: string utilization_band: type: string utilization_p95: type: number db.GPUPlacement: type: object properties: cluster: type: string cluster_id: type: string instance_id: type: string kind: type: string name: type: string namespace: type: string node_id: type: string node_name: type: string securitySchemes: BearerAuth: description: Enter "Bearer {token}" type: apiKey name: Authorization in: header ```` This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.