Head to head
Kev 4B scores 76.8 and Jev 1.14 79.3 on the System One Index. Kev 4B is more accurate on 7 of 16 text datasets, Jev 1.14 on 9. Same cases, same request, same scorer.
| Kev 4B | Jev 1.14 | |
|---|---|---|
| System One Index | 76.8 | 79.3 |
| English | 78.3 | 78.8 |
| Arabic | 74.8 | 80.3 |
| Vision Index | – | 85.1 |
| Calibration error | 5.1% | 6.0% |
| Median latency | 552 ms | 136 ms |
| Cost per 1,000 calls | $0.010 | $0.028 |
| Kev 4B | Jev 1.14 | |
|---|---|---|
| AG News English | 90.5 | 89.0 |
| SST-2 English | 92.5 | 95.0 |
| SST-5 English | 52.0 | 60.0 |
| DAIR Emotion English | 54.5 | 57.5 |
| Banking77 English | 88.0 | 78.5 |
| BoolQ English | 91.5 | 90.0 |
| XNLI English English | 90.7 | 86.0 |
| Typed decisions English | 66.3 | 74.2 |
| XNLI French French | 79.3 | 76.7 |
| XNLI Arabic Arabic | 72.7 | 71.3 |
| MASSIVE scenario Arabic | 57.0 | 65.0 |
| MASSIVE intent Arabic | 76.5 | 85.0 |
| OSACT4 offensive Arabic | 88.5 | 87.0 |
| Tweet sentiment Arabic | 63.0 | 76.0 |
| MIRACL rerank Arabic | 79.5 | 87.5 |
| MIRACL pairwise Arabic | 86.5 | 90.0 |
| Kev 4B | Jev 1.14 | |
|---|---|---|
| SST-2, as an image | – | 93.5 |
| Arabic sentiment, as an image | – | 63.5 |
| Oxford-IIIT Pet | – | 88.5 |
| VQAv2 yes/no | – | 86.5 |
| CountBench | – | 93.5 |
Model pages: Kev 4B · Jev 1.14. Latency for API models is the median per call through OpenRouter, timed from one machine.