v1.0.0·Available·
You send the policies you care about and a conversation. It returns which ones apply, and how likely each one is.
One trained model, two LoRAs.
Matcher
Which of these policies apply, and where?
bloo-match-1
Gauge
How likely is it that this one policy holds?
bloo-gauge-1
Trained model
Shared weights both LoRAs attach to.
bloobability-1-4b
Free until , then priced per token.
Input
The conversation and policies you send.
Free / 1M tokens
Output
The policies, probabilities, and evidence returned.
Free / 1M tokens
How a release ships
01
Measured, then published
Every figure on a model page is read out of the release artifacts the model was accepted on.
02
Calibration frozen at release
The map from a decision to a probability is fixed when the release ships, so cohorts are scored on one scale.
03
Trained, then held out
A candidate is trained, then scored on a holdout that was never used to fit it. Only the accepted checkpoint ships.
04
License assembled per release
We write the grant and pass through the licenses of every component the release is built on. Those upstream terms stay in force.
Access
Blooio models run behind the same account and API key as the messaging platform. The gauge always runs. The matcher joins when include_evidence is on.