---
sidebar_position: 10
title: Public Stats Feed
description: Unauthenticated network stats for Lium — USD billed, GPU-hours served and rentals per day, GPUs listed and rented per model, prices, utilisation, countries and provider reliability. The data behind lium.io/stats.
---

> ## Documentation Index
> Fetch the complete documentation index at: https://docs.lium.io/llms.txt
> Use this file to discover all available pages before exploring further.
>
> Lium (lium.io) is a GPU rental marketplace: rent NVIDIA GPUs by the hour, billed per second.

# Public Stats Feed

The figures on [lium.io/stats](https://lium.io/stats) as one unauthenticated JSON call. Aggregates only: nothing per node, per provider or per renter.

| Resource | URL |
|----------|-----|
| Live feed | [`https://lium.io/api/public/v1/stats`](https://lium.io/api/public/v1/stats) |
| Page | [`https://lium.io/stats`](https://lium.io/stats) |
| OpenAPI schema | [`https://lium.io/api/openapi.json`](https://lium.io/api/openapi.json) |

```bash
curl -sSL https://lium.io/api/public/v1/stats
```

The feed refreshes every 5 minutes; the 30-day figures (`utilisation_30d`, `median_price_30d`) update daily. Each IP may call it 60 times a minute. Fields are only ever added. A breaking change ships as `/public/v2`.

## Response

```json
{
  "generated_at": "2026-10-15T07:00:00Z",
  "fleet": {
    "gpus_listed": 1240,
    "gpus_rented": 870,
    "gpu_models": 24,
    "countries": 18,
    "providers": 95,
    "utilisation_30d": 0.71
  },
  "models": [
    {
      "gpu_model": "H200",
      "gpus_listed": 320,
      "gpus_rented": 280,
      "from_price": 2.95,
      "utilisation_30d": 0.86,
      "median_price_30d": 3.4
    }
  ],
  "countries": [{ "country": "United States", "nodes": 60 }, { "country": "Other", "nodes": 9 }],
  "reliability": { "scored_nodes": 150, "median_score": 97.5, "share_at_least_95": 0.81 },
  "utilisation_daily": [{ "day": "2026-10-14", "utilisation": 0.89 }, { "day": "2026-10-15", "utilisation": 0.91 }],
  "throughput": {
    "days": 30,
    "gmv_usd": 1560000,
    "gpu_hours": 897000,
    "rentals": 18100,
    "daily": [{ "day": "2026-10-14", "gmv_usd": 67125, "gpu_hours": 34742, "rentals": 688 }]
  }
}
```

The numbers above are an example. Read the live feed for current values.

## Fields

| Field | Meaning |
|-------|---------|
| `fleet.gpus_listed` | GPUs on verified, active nodes that take new rentals. |
| `fleet.gpus_rented` | The listed GPUs that are inside a running pod right now. |
| `fleet.countries`, `fleet.providers` | Countries and providers with at least one listed node, rented or not. |
| `fleet.utilisation_30d` | Share of the GPU-hours offered over 30 days that were rented, from 0 to 1. It covers only the models that have a `utilisation_30d`. |
| `models[].from_price` | Lowest asking price per GPU-hour, in USD, among the nodes that can be rented now. |
| `models[].utilisation_30d` | The model's 30-day rented share. `null` until the model has had one GPU-month (720 GPU-hours) on offer. |
| `models[].median_price_30d` | Median price per GPU-hour that renters paid over 30 days, weighted by GPU-hours. `null` below 10 rentals. |
| `countries` | Listed nodes, rented or not, by country: the top 10, with every country of fewer than 3 distinct providers folded into `Other`, so no row is one provider's fleet. |
| `utilisation_daily` | One row per UTC day for the last 30 days, oldest first: the share of the fleet's offered GPU-hours rented that day, 0 to 1. Today's row covers the hours so far. As with `utilisation_30d`, the fleet is the set of nodes listed now. |
| `throughput` | The last `days` complete UTC days, ending yesterday. `gmv_usd` is the USD billed to renters for pods and volumes. `gpu_hours` is the GPU time rented inside each day. `rentals` counts rentals started, excluding pods that failed to start. `daily` holds the same three figures per day, oldest first. `null` when nothing was billed in the window. |
| `reliability` | Each provider has a score over a rolling 14-day window, from 0 to 100; it is described under `reliability_score` in the [Public Nodes Feed](./public-nodes-feed.md). This object gives the median score over the listed nodes (rented or not) that have one, and the share of those nodes that score 95 or more. |

## Related

- [Public Nodes Feed](./public-nodes-feed.md): one row per node that can be rented, with its price, location and deploy link.
- [Pricing Pages and Price Feed](./pricing-feed.md): one row per GPU model, with its price range.