Self-Hosted LLM Cost Calculator

Self-Hosted Model

Model (search HuggingFace)

Infrastructure

GPU type

GPUs per node

Number of nodes

$/node/month

Other $/month

Pilot months

Workload Shape

Concurrency (sessions)

Cache hit rate (%)

Avg ISL (tokens)

Avg OSL (tokens)

Serving Performance

Output goodput (tok/s)

Input goodput (tok/s)

Output tok/s per user

Demand Estimate

Number of users

Active hours / day

Duty cycle (% of active time generating tokens)

Working days / month

Total GPUs: — Annual TCO: — Pilot: —

I:O goodput ratio: — ISL:OSL: —

Self-hosted at 100% util: —/M tokens

Monthly token demand: —

Avg utilization: — Self-hosted cost: —/M

Peak-hour utilization: —

Show derivation

Frontier API Pricing (live from OpenRouter)

Loading from OpenRouter...

Breakeven Summary

Frontier Model	Blended $/M	Breakeven Util.	Monthly Frontier Cost at Self-Hosted Volume