Head to head
Kev 4B scores 76.8 and Microsoft-Decision-1 82.0 on the System One Index. Kev 4B is more accurate on 5 of 16 text datasets, Microsoft-Decision-1 on 10, and 1 are tied. Same cases, same request, same scorer.
| Kev 4B | Microsoft-Decision-1 | |
|---|---|---|
| System One Index | 76.8 | 82.0 |
| English | 78.3 | 81.0 |
| Arabic | 74.8 | 83.2 |
| Vision Index | – | – |
| Calibration error | 5.1% | 2.8% |
| Median latency | 552 ms | 240 ms |
| Cost per 1,000 calls | $0.010 | $0.010 |
| Kev 4B | Microsoft-Decision-1 | |
|---|---|---|
| AG News English | 90.5 | 89.0 |
| SST-2 English | 92.5 | 95.5 |
| SST-5 English | 52.0 | 62.5 |
| DAIR Emotion English | 54.5 | 57.0 |
| Banking77 English | 88.0 | 87.0 |
| BoolQ English | 91.5 | 88.5 |
| XNLI English English | 90.7 | 89.3 |
| Typed decisions English | 66.3 | 79.0 |
| XNLI French French | 79.3 | 82.0 |
| XNLI Arabic Arabic | 72.7 | 72.0 |
| MASSIVE scenario Arabic | 57.0 | 83.5 |
| MASSIVE intent Arabic | 76.5 | 89.0 |
| OSACT4 offensive Arabic | 88.5 | 88.5 |
| Tweet sentiment Arabic | 63.0 | 68.5 |
| MIRACL rerank Arabic | 79.5 | 90.0 |
| MIRACL pairwise Arabic | 86.5 | 91.0 |
Model pages: Kev 4B · Microsoft-Decision-1. Latency for API models is the median per call through OpenRouter, timed from one machine.