Head to head
Naluri-279M scores 50.4 and Drex v1.5 77.6 on the System One Index. Naluri-279M is more accurate on 1 of 16 text datasets, Drex v1.5 on 15. Same cases, same request, same scorer.
| Naluri-279M | Drex v1.5 | |
|---|---|---|
| System One Index | 50.4 | 77.6 |
| English | 52.6 | 77.6 |
| Arabic | 50.7 | 77.0 |
| Vision Index | – | – |
| Calibration error | 8.3% | 10.8% |
| Median latency | – | 198 ms |
| Cost per 1,000 calls | – | $0.010 |
| Naluri-279M | Drex v1.5 | |
|---|---|---|
| AG News English | 77.5 | 90.0 |
| SST-2 English | 77.0 | 94.5 |
| SST-5 English | 34.0 | 53.5 |
| DAIR Emotion English | 68.5 | 53.5 |
| Banking77 English | 31.0 | 87.0 |
| BoolQ English | 54.5 | 89.0 |
| XNLI English English | 30.7 | 88.7 |
| Typed decisions English | 47.3 | 64.8 |
| XNLI French French | 31.3 | 82.0 |
| XNLI Arabic Arabic | 30.7 | 74.0 |
| MASSIVE scenario Arabic | 58.0 | 58.5 |
| MASSIVE intent Arabic | 52.0 | 77.5 |
| OSACT4 offensive Arabic | 82.5 | 90.5 |
| Tweet sentiment Arabic | 53.0 | 63.5 |
| MIRACL rerank Arabic | 28.0 | 89.0 |
| MIRACL pairwise Arabic | 50.5 | 86.0 |
Model pages: Naluri-279M · Drex v1.5. Latency for API models is the median per call through OpenRouter, timed from one machine.