Pending benchmarksTracking since Oct 19, 2026
Is CJRouter Clef Decision state of the art?Pending benchmarks.Announced Oct 13, 2026; no independent results yet.
CJRouter's open Qwen-based decision models (9B and 27B) with Clef-style heads that return continue/escalate/stop-style choices and calibrated probabilities.
Announced facts
The details so far
- Lab
- CJRouter
- Announced
- Oct 13, 2026
- API
- Not yet
- Open weights
- Published
- Parameters
- 9B and 27B
- Models
- qwen3.5-9b-clef-decision (Qwen3.5-9B) and qwen3.8-27b-clef-decision (Qwen3.8-27B)
- Decision head
- Clef-style joint_head with calibrated choice probabilities
- License
- Apache-2.0
- Weights
- Hugging Face under CJRouter (repos created Oct 13)
- Test set
- 400 records / 2,000 decision items (model cards)
Tags marked unsure are not stated on the lab's own pages.
Next steps
What would let us call it
Verdicts use only independent boards we already track. Two consecutive snapshots are required before any crown.
- Publicly announced with a source
- Usable by the public (API or weights)
- Listed on an independent board we track
- Enough results for a rank in a category
- Holds across two snapshots (crowns only)
Self-reported
What the lab claims
Self-reported. Shown for context; not used in any verdict.
Lab-reported scores (self-reported, own test set): 80.4% per-question accuracy (9B) and 81.4% (27B).
Receipts
Sources
LABCJRouter announcement on Xx.com/UseCJRouter/status/2112222264045318566 · Oct 19, 2026LABCJRouter weights links on Xx.com/UseCJRouter/status/2112222265483890769 · Oct 19, 2026REPOqwen3.5-9b-clef-decision on Hugging Facehuggingface.co/CJRouter/qwen3.5-9b-clef-decision · Oct 13, 2026REPOqwen3.8-27b-clef-decision on Hugging Facehuggingface.co/CJRouter/qwen3.8-27b-clef-decision · Oct 13, 2026Tracking continues until an independent board lists it.