Head to head
Kev 4B scores 76.8 and Classify-1 30.6 on the System One Index. Kev 4B is more accurate on 15 of 16 text datasets, Classify-1 on 1. Same cases, same request, same scorer.
| Kev 4B | Classify-1 | |
|---|---|---|
| System One Index | 76.8 | 30.6 |
| English | 78.3 | 30.7 |
| Arabic | 74.8 | 30.6 |
| Vision Index | – | – |
| Calibration error | 5.1% | 3.2% |
| Median latency | 552 ms | 254 ms |
| Cost per 1,000 calls | $0.010 | $0.018 |
| Kev 4B | Classify-1 | |
|---|---|---|
| AG News English | 90.5 | 26.5 |
| SST-2 English | 92.5 | 50.0 |
| SST-5 English | 52.0 | 13.0 |
| DAIR Emotion English | 54.5 | 59.0 |
| Banking77 English | 88.0 | 1.0 |
| BoolQ English | 91.5 | 38.0 |
| XNLI English English | 90.7 | 30.7 |
| Typed decisions English | 66.3 | 27.2 |
| XNLI French French | 79.3 | 30.7 |
| XNLI Arabic Arabic | 72.7 | 30.7 |
| MASSIVE scenario Arabic | 57.0 | 3.0 |
| MASSIVE intent Arabic | 76.5 | 3.5 |
| OSACT4 offensive Arabic | 88.5 | 80.0 |
| Tweet sentiment Arabic | 63.0 | 35.0 |
| MIRACL rerank Arabic | 79.5 | 12.0 |
| MIRACL pairwise Arabic | 86.5 | 50.0 |
Model pages: Kev 4B · Classify-1. Latency for API models is the median per call through OpenRouter, timed from a cloud server with 8 calls in flight.