A 144M-parameter System One model on the multilingual mmBERT-small encoder. Takes a state, a question and 2 to 20 options and returns one decision with probabilities for choice, score and yes/no questions, on a CPU.
A Gemma 4 12B fine-tune for typed decisions that also chats and reads images. Ships as GGUF for llama.cpp, holds a 64K context on a 16 GB GPU, and serves /v1/systemone next to /v1/chat/completions.
Decides
choice, score, noul, classify, route
choice, score, noul, classify, route
Architecture
julia
winnow
Fine-tuned from
jhu-clsp/mmbert-small
google/gemma-4-12b-it
License
apache-2.0
apache-2.0
Availability
Open weights
Open weights
Hosted by
—
—
Input price
—
—
Decision accuracy
73.2%
85.7%
Calibration error
—
—
Valid action rate
—
—
Median latency
—
143 ms
p95 latency
—
—
Evaluation suite
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 julia-1 and winnow?
julia-1 is from Supersonic Labs and winnow from EldanRing. Both have open weights you can download and run. Both answer choice, score, noul, classify and route questions. winnow reads up to 64K tokens of state, against 8K tokens for julia-1. julia-1 is the smaller model, at 144M parameters to 12B.
Which is more accurate, julia-1 or winnow?
They report on different suites — julia-1 73.2% on typed-decisions test set (400 cases, 2,000 questions), winnow 85.7% on JevBench public subset (231 items), Q8_0 — so the numbers do not rank them. Test both on your own labelled examples.
Which is cheaper, julia-1 or winnow?
julia-1: Free (open weights). winnow: Free (open weights). Open weights cost nothing per call beyond your own hardware.
Can I run julia-1 or winnow locally?
Yes, both: systemone pull supersonic-labs/julia-1 and systemone pull eldanring/winnow download the weights.
typed-decisions test set (400 cases, 2,000 questions)
