DecisionTune 1.0
DecisionTune's 395M decision model: ModernBERT-large fine-tuned with a 4 KB scoring head that scores a marker per option, answering a choice or yes/no question in one encoder pass with a probability for every option. Runs locally on CPU or GPU via PyTorch, MLX or ONNX.
Probabilities are a plain softmax, not calibrated, as the card says. English only; results depend on option order and how options are described. Distilled with soft labels from Cloudflare's Clef-flash. The maker's own runs of the third-party Decision Index give 29.57 on 0.2.1 (a second seed 29.13), with low Knowledge and Arts scores; the maker ran the index several times during development and designed fixes after seeing results, though no index rows were used for training. The decision-tune package offers a local /decide server, an MCP server and spreadsheet recipes. Inputs over 8,192 tokens are refused, not truncated.
What it decides
- choice — picks one option from a set
- noul — answers a yes/no question with one probability
- route — sends the input to one of several destinations
- classify — assigns a category from a fixed taxonomy
At a glance
| Parameters | 395M |
| Base model | answerdotai/ModernBERT-large |
| Maker | DecisionTune |
| Released | 2026-10-04 |
| License | apache-2.0 |
Get the weights
pip install systemonemodels
systemone pull decision-tune/decisiontune
The files are served from the maker's Hugging Face repository, decision-tune/decisiontune-1.0, and verified against the checksums recorded here.
Read more
This page was opened by System One for DecisionTune, who can claim the organisation and take it over at any time.