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.
An open family of System One models. A LoRA adapter plus a pointer head on a frozen Qwen base returns a distribution per typed question in one forward pass, serves TypeSafe's /v1/systemone contract, and ships a fitted temperature with every checkpoint.
Decides
choice, score, noul
choice, score, noul, classify, route
Architecture
djev
kev
Fine-tuned from
google/diffusiongemma-26b-a4b-it
qwen/qwen3.5-4b-base
License
apache-2.0
apache-2.0
Availability
Open weights + hosted API
Open weights
Hosted by
Maisa
—
Input price
$0.035/MTok
—
Decision accuracy
—
83.8%
Calibration error
—
0.042
Valid action rate
—
—
Median latency
—
—
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 kev?
djev is from Maisa and kev from Jared Palmer. djev has open weights and a hosted API; kev has open weights you can download and run. Both answer choice, score and noul questions. Only kev answers classify and route. kev is the smaller model, at 4.0B parameters to 26B.
Which is more accurate, djev or kev?
Only kev publishes an accuracy figure (83.8% on transfer-v4 (locked, out of domain)); djev does not, so there is no comparison to make without your own test.
Which is cheaper, djev or kev?
djev: $0.035 / $0 per 1M, or free to self-host. kev: Free (open weights). Open weights cost nothing per call beyond your own hardware.
Can I run djev or kev locally?
Yes, both: systemone pull maisa/djev and systemone pull jared-palmer/kev download the weights.
