Berget AI's System One model for gating agent commands: a LoRA adapter and a fine-tuned joint schema head on Cloudflare's Clef-Flash that answer noul, choice and score questions over a state in one forward pass. Trained on Swedish and English operations decisions.
Laya fine-tune that scans untrusted text (extracted files, knowledge-base documents, agent skills, tool descriptions) for prompt injection and data exfiltration in English and German, as a noul question answered in one forward pass. 322M parameters, PyTorch and ONNX.
Decides
noul, choice, score
noul, classify
Architecture
clef
laya
Fine-tuned from
cloudflare/clef-flash
convaiinnovations/laya-multilingual
License
apache-2.0
Unspecified; training data includes CC BY-NC-SA 4.0 material (non-commercial), not relicensed
Availability
Open weights
Open weights
Hosted by
—
—
Input price
—
—
Decision accuracy
97.0%
—
Calibration error
—
—
Valid action rate
—
—
Median latency
—
95.4 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 bev and laya-cybersec?
bev is from Berget AI and laya-cybersec from TextCortex. Both have open weights you can download and run. Both answer noul questions. Only bev answers choice and score. Only laya-cybersec answers classify. laya-cybersec is the smaller model, at 322M parameters to 9.0B. bev is licensed apache-2.0; laya-cybersec, other.
Which is more accurate, bev or laya-cybersec?
Only bev publishes an accuracy figure (97.0% on Berget held-out risk split (16,902 questions; same operations-traffic corpora as training)); laya-cybersec does not, so there is no comparison to make without your own test.
Which is cheaper, bev or laya-cybersec?
bev: Free (open weights). laya-cybersec: Free (open weights). Open weights cost nothing per call beyond your own hardware.
Can I run bev or laya-cybersec locally?
Yes, both: systemone pull berget-ai/bev and systemone pull textcortex/laya-cybersec download the weights.
—
149 ms
Evaluation suite
Berget held-out risk split (16,902 questions; same operations-traffic corpora as training)
TextCortex saved run: 279 single-window inputs, PyTorch batch one
