Head to head
d1-omni-600M scores 65.2 and Tev1 4B Experimental 71.5 on the System One Index. d1-omni-600M is more accurate on 3 of 16 text datasets, Tev1 4B Experimental on 13. Same cases, same request, same scorer.
| d1-omni-600M | Tev1 4B Experimental | |
|---|---|---|
| System One Index | 65.2 | 71.5 |
| English | 75.5 | 66.4 |
| Arabic | 53.4 | 75.8 |
| Vision Index | – | – |
| Calibration error | 9.8% | 5.3% |
| Median latency | – | 232 ms |
| Cost per 1,000 calls | – | $0.019 |
| d1-omni-600M | Tev1 4B Experimental | |
|---|---|---|
| AG News English | 87.5 | 91.5 |
| SST-2 English | 86.5 | 94.0 |
| SST-5 English | 47.0 | 43.0 |
| DAIR Emotion English | 85.0 | 56.5 |
| Banking77 English | 84.5 | 1.0 |
| BoolQ English | 74.0 | 91.0 |
| XNLI English English | 82.0 | 87.3 |
| Typed decisions English | 57.5 | 67.0 |
| XNLI French French | 65.3 | 82.7 |
| XNLI Arabic Arabic | 52.7 | 76.0 |
| MASSIVE scenario Arabic | 37.5 | 56.5 |
| MASSIVE intent Arabic | 52.0 | 70.0 |
| OSACT4 offensive Arabic | 77.5 | 84.5 |
| Tweet sentiment Arabic | 32.5 | 70.5 |
| MIRACL rerank Arabic | 55.0 | 85.5 |
| MIRACL pairwise Arabic | 66.5 | 87.5 |
Model pages: d1-omni-600M · Tev1 4B Experimental. Latency for API models is the median per call through OpenRouter, timed from one machine.