Skip to content

Scanning the GPU Market for the Best Offers

The exls CLI provides a powerful real-time view into the global GPU market. Use the offers list command to discover and compare on-demand GPU instances from dozens of providers, ensuring you always find the most cost-effective hardware for your training and inference workloads.

This guide covers how to use the command, filter results to match your needs, and interpret the output to make informed decisions.

Prerequisites

  • You’ve completed Installation.
  • You have the exls CLI available in your shell.

Supported Providers

exalsius currently aggregates offers from the following providers:

Hyperscalers

  • AWS
  • Azure
  • GCP
  • OCI
  • IBM

Neo-clouds & specialized providers

  • Lambda
  • Paperspace
  • Vast
  • Fluidstack
  • Cudo
  • Nebius
  • Hyperbolic
  • SCP
  • DigitalOcean (DO)
  • OVHcloud
  • Scaleway
  • PrimeIntellect

Note

The set of providers may evolve over time. Offers, regions, and prices are subject to change by the provider.

Command Overview

The main entrypoint is:

exls offers list --help

Output fields (typical):

  • ID – unique identifier of the offer within Exalsius
  • GPU – model (e.g., A100, H100)
  • Qty – number of GPUs in the instance
  • Cloud – provider name (e.g., AWS, LAMBDA, VAST)
  • Instance Type – instance/shape family (if applicable)
  • GPU Mem (MiB) – memory per GPU (if available)
  • vCPUs, Host Mem (MiB) – CPU and system memory of the host
  • Hourly Cost – hourly on-demand cost in USD unless noted
  • Region – region/zone/metro of the offer

Tip

You can combine multiple filters to narrow down results (e.g., GPU model + provider + price range).

Filtering Offers

Filter by model name (exact or partial match works in most shells):

exls offers list --gpu-type A100
A100 offers list

Restrict results to a specific cloud:

exls offers list --cloud-provider AWS
AWS offers list

Or combine with a GPU filter:

exls offers list --cloud-provider AWS --gpu-type A100

Show offers within a specific hourly cost window:

exls offers list --price-min 0.5 --price-max 1.0
Offers in a price range

Combine with other filters to discover best value within your budget:

exls offers list --gpu-type H100 --price-max 6

Focus on a specific GPU vendor (e.g., NVIDIA):

exls offers list --gpu-vendor NVIDIA

Tips & Best Practices

  • Combine filters to quickly narrow down to relevant offers (e.g., --gpu-type H100 --price-max 6 --cloud-provider LAMBDA).
  • Interpret “Hourly Cost” as an estimate; providers can change pricing or apply regional surcharges.
  • Mind regions: latency and data egress can materially affect total cost of ownership.
  • Check capacity/availability: some providers may have limited stock; re-run the command or select alternative regions.

Pricing & Availability

Prices shown are provider-supplied and subject to change. Always verify details with the provider before committing to a long-running workload.

Next Steps

After you've identified a promising offer, the next step is to provision your nodes.

Check out the Node Pool Guide to learn how to work with nodes!