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.
A ModernBERT-large encoder with an option-marker head. Premise and options are packed into one sequence and each option's marker is scored in a single bidirectional pass; version 1.2 makes the scoring order-invariant.
Decides
choice, score, noul, classify
choice, score, noul, classify
Architecture
hopper
von
Fine-tuned from
qwen/qwen3.5-4b
answerdotai/modernbert-large
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%
63.9%
Calibration error
0.102
0.045
Valid action rate
—
—
Median latency
—
18 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 von?
hopper is from HopitAI and von from wfzyx. Both have open weights you can download and run. Both answer choice, score, noul and classify questions. von is the smaller model, at 395M parameters to 4.0B. hopper is licensed other; von, apache-2.0.
Which is more accurate, hopper or von?
They report on different suites — hopper 68.5% on JevBench public hard tier (111 items, measured by the authors), von 63.9% on JevBench public standard tier — so the numbers do not rank them. Test both on your own labelled examples.
Which is cheaper, hopper or von?
hopper: Free (open weights). von: Free (open weights). Open weights cost nothing per call beyond your own hardware.
Can I run hopper or von locally?
Yes, both: systemone pull hopit-ai/hopper and systemone pull wfzyx/von download the weights.
JevBench public hard tier (111 items, measured by the authors)
