Teach decision models your workflow.
decide.sh models are fine-tunable and self-improving. On-premises and hosted solutions available.
For support, document review, risk and compliance.
- decide-1.181.1
- Super Decider78.5
- Myles-G 12B78.2
- kettei-fat77.7
- Jev 1.14reference70.2
Overall score out of 100 among hosted APIs, as the board shows it on October 30, 2026. decide-1.1 is listed, not yet ranked. Raw v1.8.1 results.
It learns what only your team knows.
Which team takes a ticket. Whether a scan is stamped paid. Two examples, before and after training.
A bank’s support desk
Which of the bank’s 6 teams should take this ticket?
Before training61.4%
After training89.2%
Is this ticket urgent, by the bank’s rules?
Before training86.2%
After training97%
4,000 public banking support messages, 18 minutes to train. Scored on messages it never saw.
A finance team’s scans
Is it stamped paid?
Before training, text only91.3%
After training, 150 of 150100%
Is it signed by hand?
Before training, text only50.7%
After training96%
1,500 synthetic document scans, 5.3 minutes to train. Scored on 150 it never saw.
Every fix makes it smarter.
Each new version is tested first, and goes live only when you say yes.
Fix
Send back a wrong answer with the right one. Or let it ask when it’s not sure.
Learn
It learns from your fixes, within a monthly budget you set.
Test
The new version is tested against the old one. You see what it fixed and what it broke.
Approve
It goes live when you say yes, and rolls back by itself if it gets worse.
Your fixes train only your model. Never a shared one.
Runs where your data lives.
On your servers, in your cloud, or in ours. It works the same way everywhere.
Private.
Your data trains only your model. Never ours, and never anyone else’s.
Yours to keep.
Take your model with you, along with its full report card.
Three tiers. One API.
Pay per million input tokens. Output is free.
Fast
Free to call
Routing and tagging at high volume, trained on your decisions. Through the API.
Ticket routing · Intent · Priority
Edge
$0.02 per million input tokens
More judgment, small enough for your own servers, trained on your decisions. Through the API.
Claims · Policy checks · Document review
Smart
$0.03 per million input tokens
Our strongest model, with world knowledge for the hardest calls. Through the API.
Compliance · Contracts · Fact checks
Start with decide-1.1.
Anyone can start with an API key. Want your own model? Tell us about your work.