All models

COMPARE

Put models side by side.

What each one decides, how well calibrated it is, how fast it answers and what it costs to pull. Up to 4 at a time; the better value in each row is marked.

Propertyalibi-serikbay/jevk5
SummaryQwen3.5-4B with a merged rank-16 LoRA distilled from 17,408 teacher questions and 30,000 public training rows, read out as a temperature-scaled softmax over answer-letter logits. Also in 9B, 2B and GGUF.
Decideschoice, score, noul, classify, route
Architecturejevk5
Fine-tuned fromqwen/qwen3.5-4b
Licenseapache-2.0
AvailabilityOpen weights
Hosted by—
Input price—
Decision accuracy78.4%
Calibration error0.035
Valid action rate—
Median latency13.2 ms
p95 latency—
Evaluation suiteJevBench public hard tier
Latest version0.3.0
VariantsSHA256SUMS
Size of latest version7.9 GB
Files8
Downloads0
Stars0
Tagssystem-one, qwen, lora, distilled, 4b
UpdatedSep 25, 2026