Skip to main content

Fly.io GPU Pricing

4 GPU instances across 2 regions.3 GPU models available — from $1.50/hr.

InferencePer-second billing

Developer-focused platform with GPU VMs deployable globally via anycast routing. Popular for ML inference workloads with its CLI-driven workflow and fast provisioning.

Strengths
  • Global anycast
  • Per-second billing
  • Developer-friendly CLI
  • Fast provisioning
Considerations
  • Less GPU variety
  • Not optimized for large training runs
Stop to pause billing
Egress: $0.020/GB
SSD
Quick AnswerUpdated Apr 12, 7:00 AMMethodology

Fly.io currently lists 4 GPU instances across 3 GPU models and 2 regions. Pricing starts at $1.50/hr, while the median listing price is $2.15/hr. Compare by model, commitment type, and region before treating the cheapest row as the best choice.

Starting at
$1.50/hr
Median
$2.15/hr
Models
3
Spot share
0%
Starting at
$1.50/hr
cheapest instance
GPU Models
3
available
Instances
4
total
Regions
2
covered

All Fly.io GPU Instances

4 results
GPU ModelInstanceCountVRAMRegionTypePrice/hr$/GPU/hr
A10G1x-A10G-US-East24GBUS-EastOn-Demand$1.5000Rent
A10G1x-A10G-EU-West24GBEU-WestOn-Demand$1.5000Rent
L40S1x-L40S-US-East48GBUS-EastOn-Demand$2.1500Rent
A100 80GB1x-A100-80GB-US-East80GBUS-EastOn-Demand$2.5000Rent

Fly.io GPU Cloud — FAQ

How much does Fly.io charge for GPUs?

Fly.io GPU instances start from $1.50/hr. The average price is $1.91/hr. Prices depend on GPU model, region, and commitment type (on-demand vs spot).

What GPU models does Fly.io offer?

Fly.io offers 3 GPU models: A10G, L40S, A100 80GB. Browse the full list above to compare prices per model.

Where can I see billing assumptions and risk methodology?

GPU Tracker’s pricing comparisons are paired with true cost and risk signals. Read the methodology page for how refresh cadence, cost assumptions, and reliability indicators are defined.

We use cookies for analytics and to remember your preferences. Privacy Policy