Head to head
Classify-1 scores 30.6 and d3 lite 74.9 on the System One Index. Classify-1 is more accurate on 1 of 16 text datasets, d3 lite on 15. Same cases, same request, same scorer.
| Classify-1 | d3 lite | |
|---|---|---|
| System One Index | 30.6 | 74.9 |
| English | 30.7 | 74.3 |
| Arabic | 30.6 | 74.7 |
| Vision Index | – | – |
| Calibration error | 3.2% | 7.1% |
| Median latency | 254 ms | – |
| Cost per 1,000 calls | $0.018 | – |
| Classify-1 | d3 lite | |
|---|---|---|
| AG News English | 26.5 | 85.5 |
| SST-2 English | 50.0 | 86.0 |
| SST-5 English | 13.0 | 50.5 |
| DAIR Emotion English | 59.0 | 57.0 |
| Banking77 English | 1.0 | 91.0 |
| BoolQ English | 38.0 | 81.0 |
| XNLI English English | 30.7 | 90.7 |
| Typed decisions English | 27.2 | 52.6 |
| XNLI French French | 30.7 | 81.3 |
| XNLI Arabic Arabic | 30.7 | 72.7 |
| MASSIVE scenario Arabic | 3.0 | 58.0 |
| MASSIVE intent Arabic | 3.5 | 77.5 |
| OSACT4 offensive Arabic | 80.0 | 80.5 |
| Tweet sentiment Arabic | 35.0 | 67.0 |
| MIRACL rerank Arabic | 12.0 | 82.5 |
| MIRACL pairwise Arabic | 50.0 | 84.5 |
Model pages: Classify-1 · d3 lite. Latency for API models is the median per call through OpenRouter, timed from a cloud server with 8 calls in flight.