Head to head
Tev1 4B Experimental scores 71.5 and d3 edge 65.9 on the System One Index. Tev1 4B Experimental is more accurate on 14 of 16 text datasets, d3 edge on 2. Same cases, same request, same scorer.
| Tev1 4B Experimental | d3 edge | |
|---|---|---|
| System One Index | 71.5 | 65.9 |
| English | 66.4 | 69.1 |
| Arabic | 75.8 | 62.2 |
| Vision Index | – | – |
| Calibration error | 5.3% | 18.7% |
| Median latency | 232 ms | – |
| Cost per 1,000 calls | $0.019 | – |
| Tev1 4B Experimental | d3 edge | |
|---|---|---|
| AG News English | 91.5 | 80.0 |
| SST-2 English | 94.0 | 80.5 |
| SST-5 English | 43.0 | 39.5 |
| DAIR Emotion English | 56.5 | 57.0 |
| Banking77 English | 1.0 | 90.0 |
| BoolQ English | 91.0 | 79.5 |
| XNLI English English | 87.3 | 84.0 |
| Typed decisions English | 67.0 | 42.3 |
| XNLI French French | 82.7 | 66.0 |
| XNLI Arabic Arabic | 76.0 | 59.3 |
| MASSIVE scenario Arabic | 56.5 | 41.0 |
| MASSIVE intent Arabic | 70.0 | 55.0 |
| OSACT4 offensive Arabic | 84.5 | 72.5 |
| Tweet sentiment Arabic | 70.5 | 61.5 |
| MIRACL rerank Arabic | 85.5 | 70.0 |
| MIRACL pairwise Arabic | 87.5 | 76.0 |
Model pages: Tev1 4B Experimental · d3 edge. Latency for API models is the median per call through OpenRouter, timed from one machine.