Decider
The most-downloaded open System One reproduction. Merged Qwen3.5 fine-tunes, trained on about 95 public decision sets, then calibration-aware RL and a rank-64 LoRA, that softmax letter logits at an answer slot.
This entry is decider-2b (1.9B parameters, version 11); the family also has decider-0.8b, decider-4b (v2.1), decider-35b-a3b (also in NVFP4) and decider-2b-vision, all Apache-2.0 and all serving POST /v1/systemone. On its 67-task regression set it scores 80.2% with a calibration error of 0.038; on the public hard tier of JevBench, 57.7% (decider-4b: 64.9%). Median latency is 3.2 ms with CUDA graphs, 18.9 ms eager, 6.1 ms over HTTP. A community GGUF exists.
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 | 2B |
| Base model | qwen/qwen3.5-2b-base |
| Maker | Mapika |
| Released | 2026-09-16 |
| License | apache-2.0 |
| Reported accuracy | 80.2% |
| Reported latency | 3.2 ms p50 with CUDA graphs; 18.9 ms eager; 6.1 ms p50 over HTTP |
Get the weights
pip install systemonemodels
systemone pull mapika/decider
The files are served from the maker's Hugging Face repository, Mapika/decider-2b, and verified against the checksums recorded here.
Read more
This page was opened by System One for Mapika, who can claim the organisation and take it over at any time.