~/system-one-bench / compare/nace-ai-drex-v1-5-vs-patrick-18-dexy-decision-9b
Head to head
Drex v1.5 scores 77.6 and Dexy-Decision-9B 84.7 on the System One Index. Drex v1.5 is more accurate on 1 of 16 text datasets, Dexy-Decision-9B on 15. Same cases, same request, same scorer.
| Drex v1.5 | Dexy-Decision-9B | |
|---|---|---|
| System One Index | 77.6 | 84.7 |
| English | 77.6 | 82.3 |
| Arabic | 77.0 | 86.8 |
| Vision Index | – | – |
| Calibration error | 10.8% | 3.5% |
| Median latency | 198 ms | – |
| Cost per 1,000 calls | $0.010 | – |
| Drex v1.5 | Dexy-Decision-9B | |
|---|---|---|
| AG News English | 90.0 | 91.5 |
| SST-2 English | 94.5 | 95.5 |
| SST-5 English | 53.5 | 60.0 |
| DAIR Emotion English | 53.5 | 54.0 |
| Banking77 English | 87.0 | 93.0 |
| BoolQ English | 89.0 | 89.5 |
| XNLI English English | 88.7 | 94.7 |
| Typed decisions English | 64.8 | 80.0 |
| XNLI French French | 82.0 | 89.3 |
| XNLI Arabic Arabic | 74.0 | 89.3 |
| MASSIVE scenario Arabic | 58.5 | 89.5 |
| MASSIVE intent Arabic | 77.5 | 91.5 |
| OSACT4 offensive Arabic | 90.5 | 91.5 |
| Tweet sentiment Arabic | 63.5 | 67.0 |
| MIRACL rerank Arabic | 89.0 | 88.0 |
| MIRACL pairwise Arabic | 86.0 | 91.0 |
Model pages: Drex v1.5 · Dexy-Decision-9B. Latency for API models is the median per call through OpenRouter, timed from a cloud server with 8 calls in flight.