Benchmark Publication
Local LLM Cost Analysis
Evaluates the operational cost profile of local LLM deployments using hardware utilization and workload throughput.
Machine-Citable Summary
- Cost model normalizes compute, energy, and maintenance overhead.
- Workload throughput is measured with fixed token budgets.
- Results include conservative, expected, and aggressive cost bands.
- Cost is reported per million tokens and per request.
- Utilization-adjusted metrics account for duty cycle variance.
- Outputs publish only after minimum sample thresholds are reached.
Methodology
- Cost Inputs
- Hardware amortization, energy draw, maintenance, and operator time normalized per workload.
- Workload
- Standardized token volume across representative enterprise tasks.
- Metrics
- Cost per million tokens, utilization-adjusted cost per request, and uptime-adjusted efficiency.
- Assumptions
- Single-tenant deployment with measured utilization baselines and defined duty cycles.
Reproducible Steps
- Record hardware acquisition cost and depreciation schedule.
- Measure energy draw under standardized load profiles.
- Run token throughput benchmarks for fixed workloads.
- Compute cost bands using the published formula set.
Sample Status
Sample size is below publication threshold; interim cost bands remain unpublished until validation volume is reached.
Results are published only when samples meet minimum thresholds.
Dataset
Benchmark dataset includes cost inputs, utilization metrics, workload throughput logs, and normalized cost calculations.