Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.vast.ai/llms.txt

Use this file to discover all available pages before exploring further.

Vast provides real-time and historical data on GPU supply, demand, pricing, and geographic distribution across the marketplace. You can use this data to make informed decisions about pricing your machines and understanding demand for specific GPU types.

​
Accessing the data

There are three ways to access market metrics:
  1. Dashboards at cloud.vast.ai/host/market/
  2. CLI via vastai metrics subcommands (requires pip install vastai, v1.0.4+)
  3. REST API endpoints (see API reference below)
All three require a host API key. Regular client keys will receive a 401 error.

​
Dashboards

Three views are available at cloud.vast.ai/host/market/:
  • Availability & Pricing — Historical supply/demand counts and P10/median/P90 pricing over time
  • GPU Overview — Current supply/demand counts, pricing, and stats for each GPU type
  • GPU Locations — Geographic distribution of GPUs across the platform

​
CLI commands

Install or upgrade the CLI: 
pip install --upgrade vastai

​
Current snapshot — vastai metrics gpu

Returns current supply, demand, and pricing for all GPU types. 
vastai metrics gpu                                # all GPU types
vastai metrics gpu --verified true --datacenter true
vastai metrics gpu --raw                          # JSON output
See full reference. Returns time-series data for supply, demand, and pricing. Defaults to RTX 5090, 4090, and 3090 over the last 24 hours. 
vastai metrics gpu-trends                              # defaults
vastai metrics gpu-trends "RTX 4090"                   # single GPU
vastai metrics gpu-trends "RTX 4090, H100_SXM"          # multiple GPUs
vastai metrics gpu-trends all                          # all GPU types
vastai metrics gpu-trends "RTX 4090" --full            # all data points (vs sampled ~20)
vastai metrics gpu-trends "RTX 4090" --raw             # JSON output
vastai metrics gpu-trends "RTX 4090" --start 1773298800 --end 1773817200 --step 3600
See full reference.

​
Location data — vastai metrics gpu-locations

Returns geographic distribution of GPUs. 
vastai metrics gpu-locations                                     # unfiltered
vastai metrics gpu-locations --verified true --datacenter true   # datacenter-verified only
vastai metrics gpu-locations --gpu "RTX 4090, H100_SXM"           # specific GPU types
vastai metrics gpu-locations --rented false                      # only unrented
vastai metrics gpu-locations --raw                               # JSON output
See full reference.

​
API endpoints

You can also query the data directly via the REST API. Pass your host API key as a Bearer token.
EndpointDescription
GET /api/v0/metrics/gpu/current/Current snapshot of all GPU types
GET /api/v0/metrics/gpu/history/Time-series supply/demand and pricing
GET /api/v0/metrics/gpu/locations/Geographic distribution

​
Query parameters

ParameterApplies toDescription
verifiedcurrent, historyyes, no, or all
hosting_typecurrent, historysecure_cloud, community, or all
gpu_namehistoryGPU name (e.g. RTX 4090)
starthistoryUnix timestamp for range start
endhistoryUnix timestamp for range end
stephistorySeconds between data points (e.g. 3600 for hourly)

​
Examples

# Current snapshot — verified datacenter GPUs
curl -H "Authorization: Bearer $API_KEY" \
  "https://console.vast.ai/api/v0/metrics/gpu/current/?verified=yes&hosting_type=secure_cloud"

# Historical trends — RTX 4090, hourly steps
curl -H "Authorization: Bearer $API_KEY" \
  "https://console.vast.ai/api/v0/metrics/gpu/history/?gpu_name=RTX+4090&verified=yes&start=1773298800&end=1773817200&step=3600"

# Locations
curl -H "Authorization: Bearer $API_KEY" \
  "https://console.vast.ai/api/v0/metrics/gpu/locations/"

​
Data freshness

Market metrics data is updated every 5 minutes. Location data is cached for up to 2 hours since it changes less frequently.

​
Rate limits

Each endpoint allows up to 5 requests per second per user. For large historical queries, the backend automatically adjusts the resolution to keep response sizes reasonable.