Head to head
d3 edge scores 65.9 and d3 nano 78.0 on the System One Index. d3 edge is more accurate on 0 of 16 text datasets, d3 nano on 16. Same cases, same request, same scorer.
| d3 edge | d3 nano | |
|---|---|---|
| System One Index | 65.9 | 78.0 |
| English | 69.1 | 77.9 |
| Arabic | 62.2 | 76.9 |
| Vision Index | – | – |
| Calibration error | 18.7% | 7.0% |
| Median latency | – | – |
| Cost per 1,000 calls | – | – |
| d3 edge | d3 nano | |
|---|---|---|
| AG News English | 80.0 | 88.0 |
| SST-2 English | 80.5 | 89.5 |
| SST-5 English | 39.5 | 55.5 |
| DAIR Emotion English | 57.0 | 59.5 |
| Banking77 English | 90.0 | 93.0 |
| BoolQ English | 79.5 | 88.0 |
| XNLI English English | 84.0 | 87.3 |
| Typed decisions English | 42.3 | 62.6 |
| XNLI French French | 66.0 | 86.0 |
| XNLI Arabic Arabic | 59.3 | 82.0 |
| MASSIVE scenario Arabic | 41.0 | 61.5 |
| MASSIVE intent Arabic | 55.0 | 81.0 |
| OSACT4 offensive Arabic | 72.5 | 79.0 |
| Tweet sentiment Arabic | 61.5 | 71.0 |
| MIRACL rerank Arabic | 70.0 | 77.5 |
| MIRACL pairwise Arabic | 76.0 | 86.5 |
Model pages: d3 edge · d3 nano. Latency for API models is the median per call through OpenRouter, timed from one machine.