Head to head
Microsoft-Decision-1 scores 82.0 and Classify-1 30.6 on the System One Index. Microsoft-Decision-1 is more accurate on 15 of 16 text datasets, Classify-1 on 1. Same cases, same request, same scorer.
| Microsoft-Decision-1 | Classify-1 | |
|---|---|---|
| System One Index | 82.0 | 30.6 |
| English | 81.0 | 30.7 |
| Arabic | 83.2 | 30.6 |
| Vision Index | – | – |
| Calibration error | 2.8% | 3.2% |
| Median latency | 240 ms | 254 ms |
| Cost per 1,000 calls | $0.010 | $0.018 |
| Microsoft-Decision-1 | Classify-1 | |
|---|---|---|
| AG News English | 89.0 | 26.5 |
| SST-2 English | 95.5 | 50.0 |
| SST-5 English | 62.5 | 13.0 |
| DAIR Emotion English | 57.0 | 59.0 |
| Banking77 English | 87.0 | 1.0 |
| BoolQ English | 88.5 | 38.0 |
| XNLI English English | 89.3 | 30.7 |
| Typed decisions English | 79.0 | 27.2 |
| XNLI French French | 82.0 | 30.7 |
| XNLI Arabic Arabic | 72.0 | 30.7 |
| MASSIVE scenario Arabic | 83.5 | 3.0 |
| MASSIVE intent Arabic | 89.0 | 3.5 |
| OSACT4 offensive Arabic | 88.5 | 80.0 |
| Tweet sentiment Arabic | 68.5 | 35.0 |
| MIRACL rerank Arabic | 90.0 | 12.0 |
| MIRACL pairwise Arabic | 91.0 | 50.0 |
Model pages: Microsoft-Decision-1 · Classify-1. Latency for API models is the median per call through OpenRouter, timed from a cloud server with 8 calls in flight.