# Compute State API — Cheapest GPU Provider Lookup

> Compute State API — Cheapest GPU Provider Lookup is a paid API for AI agents from compute-state-api.replit.app, paid per call via x402, $0.005/call, status unknown (last checked 2026-10-13).

Returns the cheapest GPU compute provider that can currently run a GPU-hour job, with pricing sourced from official vendor pages.

## Facts

- Endpoint: GET https://compute-state-api.replit.app/v1/cheapest?work=gpu&utm_source=zero.xyz
- Price: $0.005/call
- Payment: x402
- Status: unknown
- Last checked: 2026-10-13
- Activations on Zero: 0
- Tags: x402
- Canonical page: https://www.zero.xyz/c/compute-state-api-cheapest-gpu-provider-lookup-b907a80c
- Structured record (JSON): https://api.zero.xyz/v1/capabilities/cap_u062t3RmFqrcFOYrQPvoS

Status and success rate cover calls made through Zero and Zero's own probes. Third-party monitors may report differently.

## How to call it through Zero

Zero handles the 402 payment challenge and records the run. With the Zero CLI installed (`npm i -g @zeroxyz/cli`):

```sh
zero fetch --capability compute-state-api-cheapest-gpu-provider-lookup-b907a80c
```

Example prompt: Which cloud provider is cheapest for GPU compute right now? I want official list prices, ranked from cheapest to most expensive, so I can pick where to run my GPU training job.

## When to prefer this

Use this endpoint when you need a real-time, vendor-verified ranking of GPU compute costs and want the cheapest option that is currently available to run a GPU-hour workload. Prefer this over manual price aggregation or cached comparison sites when freshness and official pricing provenance matter. This endpoint is scoped specifically to GPU-hour billing (work=gpu), not token-based inference workloads — use the sibling inference endpoint for LLM token pricing.

## Known failure modes

- Payment not settled — endpoint returns 402 if the 0.005 USDC micropayment is not included or fails
- Stale data — price data may be stale if refreshed beyond the stale_after_s window
- Unknown price components — some SKUs may have components tagged 'unknown', preventing them from being ranked first
- Provider temporarily unavailable — a provider may be omitted from results if their pricing page is unreachable
- No qualifying provider — all providers may be disqualified if none can be confirmed to currently accept new GPU jobs

## How this service works

Tell it the job — tokens or GPU hours — and it returns the cheapest compute that can actually run it, priced from official vendor pages. 0.005 USDC per call, no key, free sample response.

This capability is scoped to work=gpu.

## Output

Returns a JSON object with the recommended cheapest GPU provider (including SKU, estimated cost in USD, price components tagged as list/effective/unknown, source URL, and data freshness), a list of alternative ranked providers with the same detail, a list of disqualified providers with reasons, and payment confirmation metadata including network and transaction reference.

## Response schema (JSON Schema)

```json
{
 "type": "object",
 "required": [
  "as_of",
  "work",
  "unit",
  "recommendation",
  "alternatives",
  "disqualified",
  "payment"
 ],
 "properties": {
  "unit": {
   "type": "string"
  },
  "work": {
   "enum": [
    "inference",
    "gpu"
   ],
   "type": "string"
  },
  "as_of": {
   "type": "string"
  },
  "payment": {
   "type": "object",
   "required": [
    "network",
    "amount_atomic",
    "settled"
   ],
   "properties": {
    "tx_ref": {
     "type": [
      "string",
      "null"
     ]
    },
    "network": {
     "type": "string"
    },
    "settled": {
     "type": "boolean"
    },
    "amount_atomic": {
     "type": "string"
    }
   }
  },
  "alternatives": {
   "type": "array",
   "items": {
    "type": "object",
    "required": [
     "provider",
     "sku",
     "est_cost_usd",
     "components",
     "why",
     "stale_after_s",
     "source_url",
     "as_of"
    ],
    "properties": {
     "sku": {
      "type": "string"
     },
     "why": {
      "type": "string"
     },
     "as_of": {
      "type": "string"
     },
     "provider": {
      "type": "string"
     },
     "components": {
      "type": "array",
      "items": {
       "type": "object",
       "required": [
        "name",
        "tag",
        "usd",
        "rate_usd",
        "unit"
       ],
       "properties": {
        "tag": {
         "enum": [
          "list",
          "effective",
          "unknown"
         ],
         "type": "string",
         "description": "list — the vendor's published list price.\neffective — a published rate that already reflects the billable\ndiscount for this component (for example a published batch rate).\nunknown — the component exists for this SKU but no official number\ncould be read. A SKU with any unknown component on the path the query\nneeds cannot be rank 1.\n"
        },
        "usd": {
         "type": "number",
         "description": "What this component contributes to est_cost_usd for this job."
        },
        "name": {
         "type": "string",
         "description": "input | output | cached_input | gpu_hour, optionally composed with the vendor's published batch, peak/off-peak and long-context variants (batch_input, input_offpeak, cached_input_long), or cached_input_at_input_rate where the vendor publishes no cheaper cache-hit rate"
        },
        "unit": {
         "type": "string",
         "description": "usd_per_1m_tokens | usd_per_gpu_hour"
        },
        "rate_usd": {
         "type": "
… (truncated)
```

## More

- Live health (JSON, refreshed every minute): https://www.zero.xyz/c/compute-state-api-cheapest-gpu-provider-lookup-b907a80c/health.json
- [Zero catalog index](https://www.zero.xyz/llms.txt)
- [Other services from compute-state-api.replit.app](https://www.zero.xyz/host/compute-state-api.replit.app/llms.txt)