01
Model APIs
OpenAI, Anthropic, Google, Mistral. Billed per million tokens.
- Batch pricing for calls that can wait
- Prompt caching for repeated context
- A smaller model for the simple calls
- Open models on a cheaper host
AI compute cost audit
Your AI bill is probably too high.We check it for free.
Send us a few months of bills for model APIs, inference or GPUs. We compare your prices with the market and get quotes from other providers. You pay 25% of what you save in the first year, and only if you switch.
Get your bills checkedWhat we look at
Most AI teams pay for compute in three ways.
Every bill has its own unit. We turn them into one price you can compare.
02
Inference platforms
Together, Fireworks, Replicate, Modal, Baseten. Billed per token or per GPU-second.
- The same model at a lower price elsewhere
- Committed spend instead of pay as you go
- Dedicated capacity when traffic is steady
03
GPU instances
AWS, Google Cloud, Azure or GPU clouds. Billed per GPU-hour.
- Reserved capacity instead of on-demand
- Independent GPU clouds instead of hyperscalers
- Interruptible GPUs for training jobs
- GPUs that are paid for but sit idle
Rough estimate
See what you could save.
Fill in what you know. Leave the rest empty.
Possible savings per year
Enter a number to see an estimate.
GPU rates are rough market prices for reserved capacity at independent GPU clouds, October 2026. API and inference ranges are estimates. Your audit gives the real number.
How it works
Two weeks from bills to quotes.
- 01
NDA, then bills
We sign your NDA first. Then you send three to six months of bills, plus usage data if you have it.
- 02
One comparable price
We break every line down to a price per GPU-hour or per million tokens and compare it with market rates.
- 03
Quotes
We ask GPU clouds and inference providers for quotes that match your workload, terms included.
- 04
Your choice
You get a short report with options, prices and switch effort. You decide. We are paid only if you switch.
The terms
- The audit is free
- Our fee is 25% of your savings in the first 12 months, based on the new contract
- No switch, no fee
- If a provider pays us a referral fee, you see it in the report
Get your bills checked
Tell us where your AI spend goes.
We reply within 24 hours with an NDA and a list of what to send.