~/system-one-bench / compare/jared-palmer-kev-0-8b-vs-meragpt-decider-1
Head to head
Kev 0.8B scores 63.6 and Decider 1 61.8 on the System One Index. Kev 0.8B is more accurate on 5 of 16 text datasets, Decider 1 on 11. Same cases, same request, same scorer.
| Kev 0.8B | Decider 1 | |
|---|---|---|
| System One Index | 63.6 | 61.8 |
| English | 71.0 | 67.4 |
| Arabic | 54.9 | 54.7 |
| Vision Index | – | – |
| Calibration error | 4.3% | 7.9% |
| Median latency | – | – |
| Cost per 1,000 calls | – | $0.012 |
| Kev 0.8B | Decider 1 | |
|---|---|---|
| AG News English | 91.0 | 87.5 |
| SST-2 English | 90.0 | 93.0 |
| SST-5 English | 47.0 | 55.5 |
| DAIR Emotion English | 56.0 | 67.5 |
| Banking77 English | 80.5 | 1.0 |
| BoolQ English | 76.0 | 80.0 |
| XNLI English English | 81.3 | 80.7 |
| Typed decisions English | 46.3 | 74.2 |
| XNLI French French | 66.0 | 67.3 |
| XNLI Arabic Arabic | 58.7 | 67.3 |
| MASSIVE scenario Arabic | 38.5 | 3.0 |
| MASSIVE intent Arabic | 51.5 | 3.5 |
| OSACT4 offensive Arabic | 49.0 | 86.0 |
| Tweet sentiment Arabic | 57.0 | 64.0 |
| MIRACL rerank Arabic | 62.5 | 81.0 |
| MIRACL pairwise Arabic | 67.0 | 78.0 |
Model pages: Kev 0.8B · Decider 1. Latency for API models is the median per call through OpenRouter, timed from a cloud server with 8 calls in flight.