Modal GPU Compute

Use this skill when configuring GPU resources, CPU/memory/disk, multi-node clusters, or handling preemption and cost trade-offs.

When to Use This Skill

• Running ML inference or training on GPUs
• Choosing between GPU types (T4, L4, A10, L40S, A100, H100, H200, B200)
• Configuring multi-GPU containers or multi-node clusters
• Configuring CPU, memory, and disk resources
• Handling preemptible/spot GPU pricing
• Optimizing GPU cost vs performance
• Multi-node distributed training
• Serving GPU inference over HTTP (cross-link modal-servers / modal-endpoints)

GPU Types Available