A local, Jev-compatible decision model from Rizzo AI Academy. XHToken's Spark-X2.5-4B with a merged typed-decisions LoRA, run on llama.cpp; yes/no, choice and score questions share one prefill of the state and are read from the answer-letter logits. No text is generated.
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, route
choice, score, noul, classify
Architecture
rizzo-flow
von
Fine-tuned from
xhtoken/spark-x2.5-4b
answerdotai/modernbert-large
License
apache-2.0
apache-2.0
Availability
Open weights
Open weights
Hosted by
—
—
Input price
—
—
Decision accuracy
64.8%
63.9%
Calibration error
0.112
0.045
Valid action rate
—
—
Median latency
195 ms
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 rizzo-flow and von?
rizzo-flow is from Rizzo AI Academy and von from wfzyx. Both have open weights you can download and run. Both answer choice, score, noul and classify questions. Only rizzo-flow answers route. von is the smaller model, at 395M parameters to 4.0B.
Which is more accurate, rizzo-flow or von?
They report on different suites — rizzo-flow 64.8% on LocalLLaMA/typed-decisions test split (400 cases, 2,000 decisions), Q8_0, 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, rizzo-flow or von?
rizzo-flow: Free (open weights). von: Free (open weights). Open weights cost nothing per call beyond your own hardware.
Which is faster, rizzo-flow or von?
By their publishers’ figures, von answers in about 18 ms at the median and rizzo-flow in about 195 ms — measured on different hardware, so treat it as a rough guide.
Can I run rizzo-flow or von locally?
Yes, both: systemone pull rizzo-ai-academy/rizzo-flow and systemone pull wfzyx/von download the weights.
LocalLLaMA/typed-decisions test split (400 cases, 2,000 decisions), Q8_0
