~/system-one-bench / compare/jericho-siahaya-naluri-279m-vs-patrick-18-dexy-decision-9b
Head to head
Naluri-279M scores 50.4 and Dexy-Decision-9B 84.7 on the System One Index. Naluri-279M is more accurate on 1 of 16 text datasets, Dexy-Decision-9B on 15. Same cases, same request, same scorer.
| Naluri-279M | Dexy-Decision-9B | |
|---|---|---|
| System One Index | 50.4 | 84.7 |
| English | 52.6 | 82.3 |
| Arabic | 50.7 | 86.8 |
| Vision Index | – | – |
| Calibration error | 8.3% | 3.5% |
| Median latency | – | – |
| Cost per 1,000 calls | – | – |
| Naluri-279M | Dexy-Decision-9B | |
|---|---|---|
| AG News English | 77.5 | 91.5 |
| SST-2 English | 77.0 | 95.5 |
| SST-5 English | 34.0 | 60.0 |
| DAIR Emotion English | 68.5 | 54.0 |
| Banking77 English | 31.0 | 93.0 |
| BoolQ English | 54.5 | 89.5 |
| XNLI English English | 30.7 | 94.7 |
| Typed decisions English | 47.3 | 80.0 |
| XNLI French French | 31.3 | 89.3 |
| XNLI Arabic Arabic | 30.7 | 89.3 |
| MASSIVE scenario Arabic | 58.0 | 89.5 |
| MASSIVE intent Arabic | 52.0 | 91.5 |
| OSACT4 offensive Arabic | 82.5 | 91.5 |
| Tweet sentiment Arabic | 53.0 | 67.0 |
| MIRACL rerank Arabic | 28.0 | 88.0 |
| MIRACL pairwise Arabic | 50.5 | 91.0 |
Model pages: Naluri-279M · Dexy-Decision-9B. Latency for API models is the median per call through OpenRouter, timed from a cloud server with 8 calls in flight.