Self-hosted vs hosted LLM inference: $/task calculator
Compare hosted token pricing against a dedicated GPU, live. Find your break-even task volume and whether self-hosting can ever beat the API at your scale.
Task profile
Self-hosted (dedicated GPU)
Hosted
$13,500.00
/mo · $0.0675/task
Self-hosted
$1,460.00
/mo · $0.0073/task
Self-hosting is cheaper by $12,040.00/mo at this volume.
- Break-even volume
- 21,630 tasks/mo
- Self-host capacity
- 788,400 tasks/mo
- Can self-hosting ever win?
- Yes
Affiliate link. We may earn a commission at no cost to you. Numbers are computed from the open pricing dataset; verify against provider quotes before committing.
Get new benchmarks first
We publish fresh, reproducible inference + agent-framework benchmarks. Subscribers get them before they’re indexed.
Frequently asked
- Is it cheaper to self-host an LLM or use a hosted API?
- It depends on volume. A dedicated GPU is a fixed monthly cost, so it only beats a per-token hosted API above a break-even task volume. Below that volume the hosted API is cheaper; above it (and within the GPU's throughput capacity) self-hosting wins. This calculator computes the exact break-even for your token profile.
- How is cost per task calculated?
- Hosted cost per task = (input tokens × input price + output tokens × output price) ÷ 1,000,000 × requests per task. Self-hosted cost per task = (GPU $/hour × hours per month) ÷ tasks per month, capped by the instance's token throughput.
- What if my GPU can't keep up?
- The tool computes the instance's monthly task capacity from its token throughput. If the break-even volume is higher than the capacity, self-hosting can never beat the API - the calculator flags this explicitly.