piqc

Overview

piqc is a free, read-only GPU waste scanner for Kubernetes inference clusters. It finds idle GPUs, oversized GPU tiers, and wasted capacity in vLLM deployments. The tool is written in Python and published as an open-source community project by Paralleliq.

Best for

Best for
Teams running vLLM on Kubernetes who want a free, safe way to spot GPU waste

Use cases

• Detect idle GPUs in Kubernetes inference clusters
• Identify oversized GPU tier allocations
• Audit vLLM deployments for wasted capacity

Notes

piqc is a free, read-only GPU waste scanner for Kubernetes inference clusters. It finds idle GPUs, oversized GPU tiers, and wasted capacity in vLLM deployments. The tool is written in Python and published as an open-source community project by Paralleliq.

32 stars on GitHub. Last updated 2026-10-04. Licensed Apache-2.0.

Use cases

• Detect idle GPUs in Kubernetes inference clusters
• Identify oversized GPU tier allocations
• Audit vLLM deployments for wasted capacity

Pros

• Read-only, so it poses no risk to running workloads
• Free and open source
• Focused specifically on GPU waste in vLLM deployments

Cons

• Limited to Kubernetes and vLLM environments
• Read-only, so it does not automatically reclaim wasted GPUs
• Small community with 32 stars, indicating early-stage adoption

Indexed from awesome-llmops and enriched against its public facts.