Superagent's 27B security decision model. It scores 2 to 16 supplied options in one forward pass and returns a probability for each, for triaging prompt injection, agent tool calls, code changes, logs and alerts. Apache-2.0 weights (gated, auto-approved) plus a hosted API.
Nandakishor M's typed-decision model. A frozen Qwen3.5 backbone is read, never trained, and cut off mid-stack; a small trained physics engine settles each choice, score or yes/no question into calibrated probabilities, a conformal set and an abstain flag. 0.8B and 4B.
Decides
choice, score, noul, classify, route
choice, score, noul, classify, route
Architecture
security-one
vega
Fine-tuned from
denis-pplx/autojev-27b
qwen/qwen3.5-0.8b
License
apache-2.0
apache-2.0
Availability
Open weights + hosted API
Open weights
Hosted by
Superagent
—
Input price
—
—
Decision accuracy
99.8%
55.6%
Calibration error
—
0.151
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 security-one and vega?
security-one is from Superagent and vega from Nandakishor M. security-one has open weights and a hosted API; vega has open weights you can download and run. Both answer choice, score, noul, classify and route questions. security-one reads up to 256K tokens of state, against 72K tokens for vega. vega is the smaller model, at 800M parameters to 27B.
Which is more accurate, security-one or vega?
They report on different suites — security-one 99.8% on BIPIA binary prompt-injection detector adaptation (800 rows; Superagent's frozen evaluation, unsafe if P(unsafe) >= 0.70), vega 55.6% on His 11 October 2026 benchmark, published with its results and its harness (a notebook in vegaml): macro accuracy over 43 public tasks for the 0.8B, 0.717 for the 4B; with JevBench (231 items) and DecisionBench (1,075 rows), 3,841 decisions in all, seed 20261011. On his own set of 2,050 choice decisions over six intent workflows (esci, clinc150, appliances, banking77, when2call, apibank), one choice question each, which he has not published with its question wording, he reports 0.763 (calibration error 0.026) for the 0.8B with its public_intents adapter, 0.389 without the adapter and 0.803 for the 4B with it. He publishes no accuracy for the typed_decisions adapters — so the numbers do not rank them. Test both on your own labelled examples.
Which is cheaper, security-one or vega?
security-one: Hosted, price not published, or free to self-host. vega: Free (open weights). Open weights cost nothing per call beyond your own hardware.
Can I run security-one or vega locally?
Yes, both: systemone pull superagent/security-one and systemone pull nandakishorm/vega download the weights.
His 11 October 2026 benchmark, published with its results and its harness (a notebook in vegaml): macro accuracy over 43 public tasks for the 0.8B, 0.717 for the 4B; with JevBench (231 items) and DecisionBench (1,075 rows), 3,841 decisions in all, seed 20261011. On his own set of 2,050 choice decisions over six intent workflows (esci, clinc150, appliances, banking77, when2call, apibank), one choice question each, which he has not published with its question wording, he reports 0.763 (calibration error 0.026) for the 0.8B with its public_intents adapter, 0.389 without the adapter and 0.803 for the 4B with it. He publishes no accuracy for the typed_decisions adapters
