A typed-decision layer for Google's DiffusionGemma, from David Villalón at Maisa AI. It compiles a request into a small answer canvas, runs one denoising read on patched vLLM and reads the probabilities of the allowed labels, for text, images and images offered as options.
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.
Decides
choice, score, noul
choice, score, noul, classify, route
Architecture
djev
rizzo-flow
Fine-tuned from
google/diffusiongemma-26b-a4b-it
xhtoken/spark-x2.5-4b
License
apache-2.0
apache-2.0
Availability
Open weights + hosted API
Open weights
Hosted by
Maisa
—
Input price
$0.035/MTok
—
Decision accuracy
—
64.8%
Calibration error
—
0.112
Valid action rate
—
—
Median latency
—
195 ms
p95 latency
—
—
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 djev and rizzo-flow?
djev is from Maisa and rizzo-flow from Rizzo AI Academy. djev has open weights and a hosted API; rizzo-flow has open weights you can download and run. Both answer choice, score and noul questions. Only rizzo-flow answers classify and route. rizzo-flow is the smaller model, at 4.0B parameters to 26B.
Which is more accurate, djev or rizzo-flow?
Only rizzo-flow publishes an accuracy figure (64.8% on LocalLLaMA/typed-decisions test split (400 cases, 2,000 decisions), Q8_0); djev does not, so there is no comparison to make without your own test.
Which is cheaper, djev or rizzo-flow?
djev: $0.035 / $0 per 1M, or free to self-host. rizzo-flow: Free (open weights). Open weights cost nothing per call beyond your own hardware.
Can I run djev or rizzo-flow locally?
Yes, both: systemone pull maisa/djev and systemone pull rizzo-ai-academy/rizzo-flow download the weights.
Evaluation suite
—
LocalLLaMA/typed-decisions test split (400 cases, 2,000 decisions), Q8_0
