Explore our compute plans and find the one that best fits your workload
Compare resources and pick the plan that fits your workload.
Explore our compute plans and find the one that best fits your workload
Set minimum VRAM, adjust CPU/RAM/storage, and a price ceiling to match the best available GPU.
Compare resources and pick the plan that fits your workload.
XS test plan for testing workloads.
From €0.10/h
current market price
Entry-level plan ideal for inference, prototyping, and small workloads. Perfect for getting started with GPU computing.
From €0.38/h
current market price
Professional plan with enhanced memory for medium-sized models and production workloads.
From €0.63/h
current market price
High-performance plan for large models, intensive training, and enterprise workloads.
From €1.69/h
current market price
Maximum performance plan for the most demanding AI training and HPC workloads. Multi-GPU configuration.
From €3.38/h
current market price
Distributed service tiers
Approved distributed profiles
0 approved profiles for this plan, with validated topologies and service classes.
Standard distributed
Private encrypted networking for development and non-critical distributed workloads.
Direct verified
Private distributed networking with validated direct connectivity before launch.
Custom requirements? Define the exact VRAM/RAM/CPU/storage you need and we will allocate the best available GPUs automatically. If we cannot satisfy your request, contact us for guaranteed capacity.
From €0.23/h
current market price
Distributed service tiers
Full cluster builder
Choose GPU, VRAM, region, nodes, GPUs per node, and service class.
Standard distributed
Private encrypted networking for development and non-critical distributed workloads.
Direct verified
Private distributed networking with validated direct connectivity before launch.
Service class
Private encrypted networking for development and non-critical distributed workloads.
Internal beta profile for experiments and flexible distributed testing.
Service class
Private distributed networking with validated direct connectivity before launch.
Use when the runtime should only start if direct connectivity has been validated first.
Direct verified
Distributed training and launches that should fail fast if direct links are not available.
Open cluster flow
