Self-Hosted vs API Break-Even Calculator
Find the volume where self-hosting beats API pricing.
With the default inputs, 2,325,581 requests/month break-even — Self-host pays off above this volume. Enter your own numbers below to recompute instantly; the full step-by-step math is shown under the worked example.
Results update automatically as you type.
- API cost / request
- $0.000645
- GPU host / month
- $1,500.00
- Break-even volume
- 2,325,581 req/mo
Self-hosting an open model has a fixed GPU cost but no per-token fee. Compare it to a hosted API to find the monthly request volume where self-hosting becomes cheaper.
How this is calculated
The API option costs (input tokens × input price + output tokens × output price) per request, scaling linearly with volume and starting at $0. Self-hosting costs a fixed monthly GPU bill regardless of volume, plus effectively no marginal per-token fee for an open model you run yourself.
The break-even volume is where the two lines cross: GPU monthly cost ÷ API cost per request = the number of monthly requests at which self-hosting becomes cheaper. Below that, the API wins; above it, self-hosting wins — assuming you can keep the GPU busy.
Is this a good result? What to do next
A low break-even (tens of thousands of requests/month) means self-hosting pays off quickly at your scale; a high break-even (millions) means the API is the pragmatic choice. But the crossover is only part of the story: self-hosting adds DevOps time, on-call, and idle-GPU waste that the raw math ignores.
Typical planning ranges
- Single mid-range cloud GPU
- ~$1,000–$2,500 / month
- GPU utilisation needed to justify self-hosting
- sustained, not bursty
- Rule of thumb
- self-host only well above break-even
Ranges are typical planning figures to sanity-check your result, not authoritative benchmarks. Your numbers will vary with use case, volume, and vendor.
How to improve this number
- Increase batch/throughput so the GPU stays busy and the effective per-request cost drops.
- Use a smaller open model or quantisation to fit cheaper hardware.
- Compare against the cheapest capable API model, not the most expensive.
Common mistakes
- Ignoring DevOps, monitoring and on-call labour in the self-hosted cost.
- Assuming 100% GPU utilisation — idle time makes self-hosting far more expensive per request.
- Forgetting that API prices keep falling, which moves the break-even against self-hosting.
When to use a different approach
If you mainly want to compare hosted models against each other, use the model cost comparison. For a specific open model's API price, use the LLM API cost calculator.
Worked example (defaults)
With the default inputs above, here is the result:
- API cost / request
- $0.000645
- GPU host / month
- $1,500.00
- Break-even volume
- 2,325,581 req/mo
- break-even = GPU monthly / API cost per request
Sources & references
Frequently asked questions
Does self-hosting really save money?+
Only above a break-even volume, and only if you can keep the GPU busy. Below that, and after adding DevOps time, a hosted API is usually cheaper and simpler.
Related calculators
Answering a real question?
This calculator powers these problem-solving guides: