A rank-16 LoRA on Qwen3.5-4B built for the JevBench setting. A document, a policy and a question go in; a calibrated distribution over the option letters comes out of one forward pass. Research and demo use only.
Qwen3.5-4B 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.
Decides
choice, score, noul, classify
choice, score, noul, classify, route
Architecture
hopper
jevk5
Fine-tuned from
qwen/qwen3.5-4b
qwen/qwen3.5-4b
License
Research and demo use only (training data includes RACE, non-commercial); serving code Apache-2.0
apache-2.0
Availability
Open weights
Open weights
Hosted by
—
—
Input price
—
—
Decision accuracy
68.5%
78.4%
Calibration error
0.102
0.035
Valid action rate
—
—
Median latency
—
13.2 ms
p95 latency
—
—
Evaluation suite
Figures are from each model’s manifest; accuracy and latency are what the publishers report, on their own suites and hardware. Add a third model.
Questions
What is the difference between hopper and jevk5?
hopper is from HopitAI and jevk5 from Alibi Serikbay. Both have open weights you can download and run. Both answer choice, score, noul and classify questions. Only jevk5 answers route. hopper is licensed other; jevk5, apache-2.0.
Which is more accurate, hopper or jevk5?
They report on different suites — hopper 68.5% on JevBench public hard tier (111 items, measured by the authors), jevk5 78.4% on JevBench public hard tier — so the numbers do not rank them. Test both on your own labelled examples.
Which is cheaper, hopper or jevk5?
hopper: Free (open weights). jevk5: Free (open weights). Open weights cost nothing per call beyond your own hardware.
Can I run hopper or jevk5 locally?
Yes, both: systemone pull hopit-ai/hopper and systemone pull alibi-serikbay/jevk5 download the weights.
JevBench public hard tier (111 items, measured by the authors)
