Typed decisions from Google's frozen Gemma-4-12B-it, with no fine-tuning. A small shim over unmodified vLLM letters the options, reads the model's own probability for each letter at one answer position and applies one calibration temperature; one output token per decision.
Frontier Infra's open decision model. A LoRA merged into Qwen3.8-27B (thinking off) that answers one typed question per forward pass with a probability over the option labels, served through AINode's TypeSafe-compatible /v1/systemone.
Decides
choice, score, noul
choice, score, noul, classify, route
Architecture
cygnet
jebadiah
Fine-tuned from
google/gemma-4-12b-it
qwen/qwen3.8-27b
License
mit
apache-2.0
Availability
Open weights
Open weights
Hosted by
—
—
Input price
—
—
Decision accuracy
87.9%
86.6%
Calibration error
—
0.113
Valid action rate
—
—
Median latency
50 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 cygnet and jebadiah?
cygnet is from Blockbrain Labs and jebadiah from Frontier Infra. Both have open weights you can download and run. Both answer choice, score and noul questions. Only jebadiah answers classify and route. cygnet is the smaller model, at 12B parameters to 27B. cygnet is licensed mit; jebadiah, apache-2.0.
Which is more accurate, cygnet or jebadiah?
They report on different suites — cygnet 87.9% on JevBench public set (231 items), JevBench CLI, jebadiah 86.6% on JevBench public set (231 items), maker's run with the v1.4.2 harness; ECE on the hard tier — so the numbers do not rank them. Test both on your own labelled examples.
Which is cheaper, cygnet or jebadiah?
cygnet: Free (open weights). jebadiah: Free (open weights). Open weights cost nothing per call beyond your own hardware.
Can I run cygnet or jebadiah locally?
Yes, both: systemone pull blockbrain-labs/cygnet and systemone pull frontier-infra/jebadiah download the weights.
Evaluation suite
JevBench public set (231 items), JevBench CLI
JevBench public set (231 items), maker's run with the v1.4.2 harness; ECE on the hard tier
