NOMEDA-Decision v0.1-beta
An Arabic-first decision model: a 12.8M-parameter decision layer over a frozen SILMA Arabic embedding encoder answers many choice, score and yes/no questions about one state at once, with calibrated probabilities and an abstain signal. Early beta.
The encoder is not bundled; the runtime (the nomeda pip package) pulls it at a pinned revision. The maker reports NLI and commonsense at chance level and offensive-language detection below the most-common-answer baseline; 86.9% of the test set comes from task families seen in training. Dataset licences are not listed, despite the card's reference to a licence file.
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
- route — sends the input to one of several destinations
At a glance
| Parameters | 148M |
| Base model | silma-ai/silma-embedding-matryoshka-v0.1 |
| Maker | NOMEDA |
| Released | 2026-10-01 |
| License | apache-2.0 |
| Reported accuracy | 77.2% |
| Reported latency | 92 ms p50 / 187 ms p95 on one A6000 |
Get the weights
pip install systemonemodels
systemone pull nomeda-lab/nomeda-decision
The files are served from the maker's Hugging Face repository, nomeda-lab/nomeda-decision-v0.1-beta, and verified against the checksums recorded here.
Read more
This page was opened by System One for NOMEDA, who can claim the organisation and take it over at any time.