System Requirements

Sizing guidance for on-prem deployment profiles.

Sizing values below are planning baselines for integrator workshops. They are indicative and must be validated in customer pre-production tests. The runtime sizing source of truth is ../Noxa/docs/system-requirements.md. As of 2026-04-04, no public NOXA benchmark report is published with guaranteed throughput/latency numbers.

Sizing Matrix

Minimum, recommended, and optimal baselines

Each tier is shown for deployments without local AI and with optional local AI enabled.

Minimum
Without local AI
2 vCPU
4 GB RAM
40 GB SSD
Demo/Lab or controlled pilot baseline
With local AI
12 vCPU
32 GB RAM
800 GB SSD/NVMe
Local AI enabled baseline from runtime sizing guide

Matches runtime minimum planning baseline for demo/lab and initial validation.

Recommended
Without local AI
8 vCPU
16 GB RAM
250 GB SSD/NVMe
Standard customer profile with daily backup discipline
With local AI
12-16 vCPU
32-64 GB RAM
800 GB+ NVMe
Optional GPU based on selected local model footprint

Default production onboarding baseline for standard on-prem workloads.

Optimal
Without local AI
12 vCPU+
24 GB RAM+
500 GB+ NVMe
Comfortable production headroom with stronger backup/monitoring margins
With local AI
16 vCPU+
64 GB RAM+
1 TB+ NVMe
GPU optional but recommended for larger local model tiers

Used for multi-team sustained loads and stronger growth headroom.

Profile Mapping

How profiles map to sizing tiers

Demo / Lab
Minimum

Fast setup for validation and walkthroughs.

Single Server
Minimum to Recommended

Starts from small-customer baseline and scales to recommended production headroom.

Standard / Hardened On-Prem
Recommended to Optimal

Depends on concurrency, compliance controls, and retention policy.

AI-Enabled
Recommended AI to Optimal AI

Local model choice is the largest sizing driver.

Need a sizing workshop with your integrator and platform team?