What will your AI workload actually cost?
Which open-weight models will run on your hardware?
A calculator that compares API pricing across providers for real jobs. Assumptions are shown, and every price links to the provider's own pricing page with the date it was last checked.
Pick a GPU or a Mac and see which models fit, at which quantization and context length. The memory math is shown for every result, with a link to each model's published specs. Fit is calculated, not benchmarked.
- 1,000 support tickets triaged
- 10,000 documents summarized
- A coding agent running all day
- Your own token volumes
- A single consumer GPU
- A Mac with unified memory
- A multi-GPU workstation
- Your own memory budget
What a sponsor gets
One labeled placement at the top of every comparison result. It is marked "Sponsored" and never changes the calculated costs or the order of results.
A labeled "Run it on your product" placement beside every model that fits the visitor's hardware. It is marked "Sponsored" and never changes the fit calculation.
How it will work
- Pick a workload or enter your own token volumes.
- See cost, rate limits, and latency tradeoffs side by side.
- Follow any number to the provider's pricing page it came from.
- Choose your GPU or Mac, or enter your memory budget.
- See which models fit, with the memory math for each.
- Follow any spec to the source it came from.
Where this stands
The site is not live and has no visitors yet, so there are no numbers to quote. Once there are real visit counts, sponsors will get them.
Taking a founding slot is a written yes by email. No money changes hands until the site exists.
Take the slot or tell me no
If a founding slot fits, reply to my message or write to todd@toddeidt.com. A plain "no" is a useful answer too.
If a founding listing fits, reply to my message or write to todd@toddeidt.com. A plain "no" is a useful answer too.