DECISION & CLASSIFICATION / Jared Palmer
Kev-4B
An open decision model with a pointer head, typed probabilities and a Jev-compatible serving interface.
ACCESS & CAPABILITIES
How it fits your workflow
Self-host on an NVIDIA GPU or Apple Silicon via MLX. Training and serving code are available; CPU deployment is not verified here.
- Documented judgment types
- Choose an option · Yes / no probability · Rubric score
- Languages
- English; the model card places other languages outside its intended scope.
- Context and input limits
- Validated context: 8,192 tokens. The server accepts longer states, but that is not equivalent to validated accuracy.
- Jev-compatible interface
- Documented; not an accuracy guarantee
PRICE & PROBABILITIES
Hosted price unverified
No hosted rate verified for this entry. Self-hosting needs compute, capacity and maintenance; downloadable weights do not make inference free.
Interpret the output
The release applies a fitted temperature. Probability thresholds need rechecking after model, domain or calibration changes.
Check before choosing
API compatibility does not establish equal accuracy or transferable thresholds.
Keep the checkpoint and precision with any benchmark; this catalog does not combine its different evaluation suites.
No shared benchmark measured yet
Our accuracy, calibration, p50/p95 latency and cost per thousand decisions are not measured. This is a source-reviewed capability profile. Compare the same task, model version, language, input size and serving setup before ranking results.
Read the selection and evaluation guide →Official sources & version notes
Jared Palmer · Kev-4B model card ↗ Checked 2026-10-05