Observed arrival · 2026-09-20
openinstinct, an open decision model
An in-development model that turns text, questions, and answer choices into probabilities for routing or other software decisions.
- For
- Developers building routing and decision workflows
- Worth noticing
- The Phase 2 checkpoint has 1.7 billion parameters and scores one question per pass.
Field notes
The project frames inference as a probability-producing decision layer rather than a text-generation interface. Its documented inputs include support messages, alerts, documents, and JSON, while the output is a score for each application-defined option. The homepage reports a 1.7-billion-parameter checkpoint and roughly 98 ms median latency from a Phase 2 evaluation run, but notes that calibration remains unfinished and the weights are not yet available.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
One card from the complete issue