Cluster vRAM Production Profiler

vRAM Profiler

Total Cluster vRAM

0 GB

Model Weights

0 GB

Pre-Allocated KV Pool

0 GB

Active Consumption

0 GB

Max Theoretical Concurrency

0 Users

GPU vRAM Per Card

GPU Count (Tensor Parallelism)

GPU Utilisation Cap

Model Weights

Active Concurrent Requests

* Note: This profiler represents the absolute mathematical memory ceiling. Real-world concurrency will also be constrained by compute bounds, network overhead, and continuous batching efficiency.