circuit-1.7b
James Barney's open System One models: a LoRA adapter plus a pointer readout head on Qwen base models that answers noul, choice and score questions with a probability per option in one pass. Text (1.7B, 8B), image (VL-4B) and audio (7B) variants, plus a free hosted API.
Adapter and head, not a standalone model; it needs the circuit repo's scorer, which applies a parallel-options mask (about 5 points lower without it). v2.0 and later add multi (every option that applies) and locate questions on text states. Siblings circuit-8b, circuit-vl-4b and circuit-audio-7b. The maker's caveats: the cold-eval train splits are in the training mix (82% MultiNLI, 97% SMS spam, 91% on 151-way CLINC for this size); confidence on undecidable inputs is not reliable; English only; intended for research and as a decision-circuits backend, not for decisions that affect people. The hosted endpoint is free and rate-limited, with no uptime commitment, and may lag the latest revision. head.pt is a PyTorch pickle.
What it decides
- choice — picks one option from a set
- score — places the input on an ordered scale
- noul — answers a yes/no question with one probability
- classify — assigns a category from a fixed taxonomy
At a glance
| Parameters | 1.7B |
| Base model | Qwen/Qwen3-1.7B-Base |
| Maker | James Barney |
| Released | 2026-09-19 |
| License | apache-2.0 |
Hosted API
Served by Decision Circuits — $0/MTok input, $0/MTok output. Get access.
Get the weights
pip install systemonemodels
systemone pull james-barney/circuit
The files are served from the maker's Hugging Face repository, jbarney/circuit-1.7b, and verified against the checksums recorded here.
Read more
This page was opened by System One for James Barney, who can claim the organisation and take it over at any time.