Instance Types

Executing a run on Spell requires specifying the instance type which you would like to use for the run. For example, the spell run CLI command accepts a machine-type flag for this purpose:

spell run --machine-type K80 'python example.py'

Here's a list of all of the instance types available in Spell Community. Note that Spell is free to use with the basic CPU instance type—no credit card required.

Note

Users on Spell for Teams using our AWS or GCP cluster integrations can define and use their own machine types, including additional instance types not available in Community. See the page Cluster Machine Type Management for more details.

CPU instances

Machine Type vCPUs RAM (GB) Price $/hour Price $/second
cpu (default) 2 4 Free Free
cpu-big 16 32 $0.68 $0.00019
cpu-huge 72 144 $3.06 $0.00085
ram-big 16 128 $1.01 $0.00028
ram-huge 96 768 $6.05 $0.00168

GPU instances (K80)

Machine Type NVIDIA GPU VRAM (GB) TFlops vCPUs RAM (GB) Price $/hour Price $/second
K80 1 x Tesla K80 12.0 4.3 4 61 $0.90 $0.00025
K80x2 2 x Tesla K80 24.0 8.6 8 122 $1.80 $0.00050
K80x4 4 x Tesla K80 48.0 17.2 16 244 $3.60 $0.00100
K80x8 8 x Tesla K80 96.0 34.4 32 488 $7.20 $0.00200

GPU Instances (V100)

Machine Type NVIDIA GPU VRAM (GB) TFlops vCPUs RAM (GB) Price $/hour Price $/second
V100 1 x Tesla V100 16.0 15.7 8 61 $3.06 $0.00085
V100x2 2 x Tesla V100 32.0 31.4 16 122 $6.12 $0.00170
V100x4 4 x Tesla V100 64.0 62.8 32 244 $12.24 $0.00340
V100x8 8 x Tesla V100 128.0 125.6 64 488 $24.48 $0.00680