Observed arrival · 2026-09-21
Antares Labs’ Hunch, a model for judgment before language
An independent AI lab’s research preview for classifying situations by returning probabilities over candidate answers in one forward pass.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- Software builders working on routing, triage, or decision workflows
- Worth noticing
- The preview distinguishes Choice, Boolean, and Score questions and illustrates ticket routing with Billing at 0.91.
Field notes
Hunch frames prediction as a probability distribution over application-defined candidates rather than a generated response. Its displayed example routes a duplicate-charge support ticket among Billing, Account, and Technical support, while the surrounding research note emphasizes that calibration may vary by task. The site presents the model as a research preview and says evaluation, unfamiliar task families, and release materials remain in progress.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
✦PrettyNotable craft visible
●ProPolished or operationally mature
◎NicheUnusually specific use
One card from the complete issue