Skip to content

Command Palette

Search for a command to run...

Plan & Cost

API vs Local

Compare paying per token with running your own models, including hardware write-off, operations and power, and see where self-hosting breaks even.

Start from:

Today's API spend

Estimate it in the Cost Calculator

Self-hosting

API per month

$1,500

Self-hosted per month

$1,802

Self-hosting costs more

$302

per month

Where self-hosting breaks even

Self-hosting becomes cheaper once your API bill passes $1,802 a month, about 120% of today's volume.

Before deciding on cost alone

Self-hosted open-weight models may not match the quality, latency or availability of the API model you use today. Run your own evaluation set on both, and check data-residency and compliance needs. Then find models that fit your hardware in the Local Model Advisor.

Self-hosting costs more per month

$302

See results
Assumptions & sources

Self-hosted monthly cost = hardware ÷ write-off months + cloud GPU or hosting + operations + electricity. It is treated as flat until the hardware runs out of capacity; API cost is assumed to scale with usage.

Presets are illustrative setups in USD, not quotes; replace them with your own prices. Operations is usually the cost people forget: someone has to patch, monitor and upgrade the system.

Estimate only. It does not imply equivalent quality, availability, latency or compliance.