GPU Availability Guide | Akash Network - Your Guide to Decentralized Cloud

GPU availability data is embedded in provider responses — there is no standalone GPU endpoint. This guide shows how to query providers and filter for specific GPU resources using the Providers API.

Base URL: https://console-api.akash.network

Finding GPU Providers

Step 1: Fetch All Providers

curl https://console-api.akash.network/v1/providers

Each provider in the response includes a gpuModels array:

{
  "owner": "akash1u5cdg7k3gl43mukca4aeultuz8x2j68mgwn28e",
  "isOnline": true,
  "gpuModels": [
    {
      "vendor": "nvidia",
      "model": "rtx4060ti",
      "ram": "16Gi",
      "interface": "PCIe"
    }
  ],
  "stats": {
    "gpu": { "active": 0, "available": 1, "pending": 0, "total": 1 }
  }
}

Step 2: Filter for Online Providers with GPUs

const response = await fetch("https://console-api.akash.network/v1/providers");
const providers = await response.json();

const gpuProviders = providers.filter(
  (p) => p.isOnline && p.gpuModels.length > 0
);

Step 3: Filter by GPU Vendor

const nvidiaProviders = gpuProviders.filter((p) =>
  p.gpuModels.some((gpu) => gpu.vendor === "nvidia")
);

Step 4: Filter by Specific GPU Model

const t4Providers = gpuProviders.filter((p) =>
  p.gpuModels.some((gpu) => gpu.model === "t4")
);

Step 5: Filter by GPU Memory

const highMemProviders = gpuProviders.filter((p) =>
  p.gpuModels.some((gpu) => gpu.ram === "80Gi")
);

Checking GPU Availability

The stats.gpu field on each provider shows real-time GPU capacity:

Field	Description
`stats.gpu.total`	Total GPU units on the provider
`stats.gpu.active`	Currently leased GPUs
`stats.gpu.available`	GPUs available for new deployments
`stats.gpu.pending`	GPUs in pending state

Filter for providers with available GPUs:

const availableGpuProviders = gpuProviders.filter(
  (p) => p.stats.gpu.available > 0
);

Complete Example

This script finds all online providers with available NVIDIA GPUs and prints their details:

async function findAvailableGpuProviders(vendor, model) {
  const response = await fetch("https://console-api.akash.network/v1/providers");
  const providers = await response.json();

  return providers
    .filter((p) => {
      if (!p.isOnline || p.gpuModels.length === 0) return false;
      if (p.stats.gpu.available === 0) return false;

      return p.gpuModels.some((gpu) => {
        const vendorMatch = !vendor || gpu.vendor === vendor;
        const modelMatch = !model || gpu.model === model;
        return vendorMatch && modelMatch;
      });
    })
    .map((p) => ({
      owner: p.owner,
      hostUri: p.hostUri,
      gpuModels: p.gpuModels,
      availableGpus: p.stats.gpu.available,
      totalGpus: p.stats.gpu.total,
    }));
}

// Find all providers with available NVIDIA GPUs
const providers = await findAvailableGpuProviders("nvidia");
console.log(`Found ${providers.length} providers with available NVIDIA GPUs:`);
providers.forEach((p) => {
  console.log(`  ${p.owner} — ${p.availableGpus}/${p.totalGpus} GPUs available`);
  p.gpuModels.forEach((gpu) => {
    console.log(`    ${gpu.vendor} ${gpu.model} (${gpu.ram}, ${gpu.interface})`);
  });
});

// Find providers with available T4 GPUs specifically
const t4Providers = await findAvailableGpuProviders("nvidia", "t4");
console.log(`\nFound ${t4Providers.length} providers with available T4 GPUs`);

GPU Model Fields Reference

Each entry in the gpuModels array contains:

Field	Type	Description	Examples
`vendor`	string	GPU manufacturer	`"nvidia"`
`model`	string	GPU model identifier	`"rtx4060ti"`, `"t4"`, `"a100"`, `"h100"`
`ram`	string	GPU memory	`"16Gi"`, `"40Gi"`, `"80Gi"`
`interface`	string	Connection interface	`"PCIe"`, `"SXM"`

See also: Providers API Reference for the complete provider response schema.