Jebadiah 27B
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.
Trained on public data only (LocalLLaMA/typed-decisions, BoolQ, MNLI, DBpedia-14, HelpSteer2, SummEval). The maker's own run on JevBench's 231 public items gives 86.6% (200 of 231; 82 of 111 hard) with hard-tier ECE 0.113, and a 78.9% macro over six public suites. The Decision Index 0.2.1 maintainer validated its run at 54.67, fifth of 67 open models on 26 September. Siblings 9B v2 and 4B v2 use the same recipe on Qwen3.5; GGUF and MLX builds are published.
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 calibrated probability
- classify — assigns a category from a fixed taxonomy
- route — sends the input to one of several destinations
At a glance
| Parameters | 27B |
| Base model | qwen/qwen3.8-27b |
| Maker | Frontier Infra |
| Released | 2026-09-26 |
| License | apache-2.0 |
| Reported accuracy | 86.6% |
Get the weights
pip install systemonemodels
systemone pull frontier-infra/jebadiah
The files are served from the maker's Hugging Face repository, frontier-infra/jebadiah-27b, and verified against the checksums recorded here.
Read more
This page was opened by System One for Frontier Infra, who can claim the organisation and take it over at any time.