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.
meraGPT's hosted decision model (state-decider-1). Answers yes/no, choice and rubric-score questions over one state as calibrated distributions in a single pass, on the System One schema, so the typesafe-sdk works by changing its base URL.
Decides
choice, score, noul
choice, score, noul, classify, route
Architecture
cygnet
state-decider
Fine-tuned from
google/gemma-4-12b-it
—
License
mit
proprietary
Availability
Open weights
Hosted API
Hosted by
—
meraGPT
Input price
—
$0.030/MTok
Decision accuracy
87.9%
76.8%
Calibration error
—
—
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 state-decider-1?
cygnet is from Blockbrain Labs and state-decider-1 from meraGPT. cygnet has open weights you can download and run; state-decider-1 is only available as a hosted API. Both answer choice, score and noul questions. Only state-decider-1 answers classify and route. cygnet reads up to 16K tokens of state, against 4K tokens for state-decider-1. cygnet is licensed mit; state-decider-1, proprietary.
Which is more accurate, cygnet or state-decider-1?
They report on different suites — cygnet 87.9% on JevBench public set (231 items), JevBench CLI, state-decider-1 76.8% on typed-decisions benchmark (400 cases, 2,000 decisions; teacher-ensemble references), single run — so the numbers do not rank them. Test both on your own labelled examples.
Which is cheaper, cygnet or state-decider-1?
cygnet: Free (open weights). state-decider-1: $0.03 / $0 per 1M. Open weights cost nothing per call beyond your own hardware.
Can I run cygnet or state-decider-1 locally?
cygnet yes — systemone pull blockbrain-labs/cygnet downloads its weights. The other is only served as a hosted API.
Evaluation suite
JevBench public set (231 items), JevBench CLI
typed-decisions benchmark (400 cases, 2,000 decisions; teacher-ensemble references), single run
