Head to head
Drex v1.5 scores 77.6 and Tev1 4B Experimental 71.5 on the System One Index. Drex v1.5 is more accurate on 8 of 16 text datasets, Tev1 4B Experimental on 8. Same cases, same request, same scorer.
| Drex v1.5 | Tev1 4B Experimental | |
|---|---|---|
| System One Index | 77.6 | 71.5 |
| English | 77.6 | 66.4 |
| Arabic | 77.0 | 75.8 |
| Vision Index | – | – |
| Calibration error | 10.8% | 5.3% |
| Median latency | 198 ms | 232 ms |
| Cost per 1,000 calls | $0.010 | $0.019 |
| Drex v1.5 | Tev1 4B Experimental | |
|---|---|---|
| AG News English | 90.0 | 91.5 |
| SST-2 English | 94.5 | 94.0 |
| SST-5 English | 53.5 | 43.0 |
| DAIR Emotion English | 53.5 | 56.5 |
| Banking77 English | 87.0 | 1.0 |
| BoolQ English | 89.0 | 91.0 |
| XNLI English English | 88.7 | 87.3 |
| Typed decisions English | 64.8 | 67.0 |
| XNLI French French | 82.0 | 82.7 |
| XNLI Arabic Arabic | 74.0 | 76.0 |
| MASSIVE scenario Arabic | 58.5 | 56.5 |
| MASSIVE intent Arabic | 77.5 | 70.0 |
| OSACT4 offensive Arabic | 90.5 | 84.5 |
| Tweet sentiment Arabic | 63.5 | 70.5 |
| MIRACL rerank Arabic | 89.0 | 85.5 |
| MIRACL pairwise Arabic | 86.0 | 87.5 |
Model pages: Drex v1.5 · Tev1 4B Experimental. Latency for API models is the median per call through OpenRouter, timed from one machine.