One of the earliest Jev-shaped scorers. A sequence-classification head on Qwen3.5-4B-Base scores each (state, question, option) triple and softmaxes per question. Non-commercial licence.
A Gemma 4 12B fine-tune for typed decisions that also chats and reads images. Ships as GGUF for llama.cpp, holds a 64K context on a 16 GB GPU, and serves /v1/systemone next to /v1/chat/completions.
Decides
choice, score, noul, classify
choice, score, noul, classify, route
Architecture
scorer
winnow
Fine-tuned from
qwen/qwen3.5-4b-base
google/gemma-4-12b-it
License
cc-by-nc-4.0
apache-2.0
Availability
Open weights
Open weights
Hosted by
—
—
Input price
—
—
Decision accuracy
70.7%
85.7%
Calibration error
0.044
—
Valid action rate
—
—
Median latency
112 ms
143 ms
p95 latency
—
—
Evaluation suite
author-reported
JevBench public subset (231 items), Q8_0
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 system-one-qwen3-5-4b-scorer and winnow?
system-one-qwen3-5-4b-scorer is from pngwn and winnow from EldanRing. Both have open weights you can download and run. Both answer choice, score, noul and classify questions. Only winnow answers route. system-one-qwen3-5-4b-scorer is the smaller model, at 4.0B parameters to 12B. system-one-qwen3-5-4b-scorer is licensed cc-by-nc-4.0; winnow, apache-2.0.
Which is more accurate, system-one-qwen3-5-4b-scorer or winnow?
They report on different suites — system-one-qwen3-5-4b-scorer 70.7% on author-reported, winnow 85.7% on JevBench public subset (231 items), Q8_0 — so the numbers do not rank them. Test both on your own labelled examples.
Which is cheaper, system-one-qwen3-5-4b-scorer or winnow?
system-one-qwen3-5-4b-scorer: Free (open weights). winnow: Free (open weights). Open weights cost nothing per call beyond your own hardware.
Which is faster, system-one-qwen3-5-4b-scorer or winnow?
By their publishers’ figures, system-one-qwen3-5-4b-scorer answers in about 112 ms at the median and winnow in about 143 ms — measured on different hardware, so treat it as a rough guide.
Can I run system-one-qwen3-5-4b-scorer or winnow locally?
Yes, both: systemone pull pngwn/system-one-qwen3-5-4b-scorer and systemone pull eldanring/winnow download the weights.