HEAD TO HEAD
Yoav Pinto: nativ-he-decision and StartLux: startlux-decision, compared on what they decide, where they run, what they cost and what their publishers report.
| Property | yoav-pinto/nativ-he-decision | startlux/startlux-decision |
|---|---|---|
| Summary | A 364M Hebrew decision encoder on NeoDictaBERT: give it a state, a question and 2 to 7 free-text options and it returns a probability for each option in one forward pass, about 70 ms on a laptop CPU. Hebrew only. | StartLux's decision family in six sizes, 0.8B to a 35B-A3B mixture of experts. All questions in a request are answered in one forward pass, with a probability for every option, through a TypeSafe /v1/systemone-compatible server. Text, JSON or images, up to 256K tokens. |
| Decides | choice, classify, route | choice, score, noul |
| Architecture | nativ | startlux-decision |
| Fine-tuned from | dicta-il/neodictabert | — |
| License | cc-by-4.0 | cc-by-nc-4.0 |
| Availability | Open weights | Open weights |
| Hosted by | — | — |
| Input price | — | — |
| Decision accuracy | 96.2% | 88.3% |
| Calibration error | 0.010 | — |
| Valid action rate | — | — |
| Median latency | — | 26 ms |
| p95 latency | — | — |
| Evaluation suite | nativ-bench user-intent task, 4 options (MASSIVE Hebrew, 2,973 items; the maker's own suite) | JevBench public set (231 items; 204 correct), maker's run |
| Latest version | 2026.09 | 2026.10 |
| Variants | — | LICENSE, NOTICE, startlux_decision |
| Size of latest version | 697.3 MB | 8.7 GB |
| Files | 5 | 23 |
| Downloads | 0 | 3 |
| Stars | 0 | 0 |
| Tags | system-one, hebrew, encoder, cpu, 364m | system-one, qwen, non-commercial, gguf, cuda-graphs, vision, mlx, 4.7b |
| Updated | Oct 7, 2026 | Oct 7, 2026 |
Figures are from each model’s manifest; accuracy and latency are what the publishers report, on their own suites and hardware. Add a third model.
nativ-he-decision is from Yoav Pinto and startlux-decision from StartLux. Both have open weights you can download and run. Both answer choice questions. Only nativ-he-decision answers classify and route. Only startlux-decision answers score and noul. startlux-decision reads up to 256K tokens of state, against 1K tokens for nativ-he-decision. nativ-he-decision is the smaller model, at 364M parameters to 4.7B. nativ-he-decision is licensed cc-by-4.0; startlux-decision, cc-by-nc-4.0.
They report on different suites — nativ-he-decision 96.2% on nativ-bench user-intent task, 4 options (MASSIVE Hebrew, 2,973 items; the maker's own suite), startlux-decision 88.3% on JevBench public set (231 items; 204 correct), maker's run — so the numbers do not rank them. Test both on your own labelled examples.
nativ-he-decision: Free (open weights). startlux-decision: Free (open weights). Open weights cost nothing per call beyond your own hardware.
Yes, both: systemone pull yoav-pinto/nativ-he-decision and systemone pull startlux/startlux-decision download the weights.