The same work, for less.
Smaller models need less compute, and we pass that on. We bill per turn, so you know what a conversation costs before it starts.
Long conversations don’t get more expensive
Most models process the whole conversation again on every turn. The longer it runs, the more each turn costs.
tbd-0 works with a context of constant size. Turn 50 costs the same as turn 1.
Where else the savings come from
-
Smaller models
Instead of one giant model for everything, tbd-0 uses models sized for the job. Less compute per answer means a lower price.
-
Expert models
For work in a specific field, a small expert model outperforms a large general-purpose one. You don’t pay for knowledge you don’t need.
-
Per-turn billing
We bill per model turn, not per token. Long prompts and long conversations don’t lead to surprises on the invoice.
-
No switching cost
Our API is compatible with the OpenAI Responses API. Testing tbd-0 against your current setup takes one line.
See what it would cost you.
Tell us about your use case and volume, and we’ll come back with a concrete price.