Jared Palmer · OpenRouter API
Kev 4B by Jared Palmer scores 76.8 on the System One Index (13th of 24). Calibration error 5.1%, median latency 552 ms.
| Dataset | Accuracy | Rank | Best on the board | Calibration error |
|---|---|---|---|---|
| AG News English | 90.5 | 6th of 24 | 92.5 | 3.4% |
| SST-2 English | 92.5 | 14th of 24 | 97.0 | 4.5% |
| SST-5 English | 52.0 | 12th of 24 | 62.5 | 13.9% |
| DAIR Emotion English | 54.5 | 20th of 24 | 91.5 | 12.3% |
| Banking77 English | 88.0 | 8th of 24 | 98.5 | 12.2% |
| BoolQ English | 91.5 | 2nd of 24 | 92.5 | 2.1% |
| XNLI English English | 90.7 | 4th of 24 | 93.3 | 6.7% |
| Typed decisions English | 66.3 | 13th of 24 | 79.0 | 9.0% |
| XNLI French French | 79.3 | 12th of 24 | 90.7 | 14.1% |
| XNLI Arabic Arabic | 72.7 | 11th of 26 | 89.3 | 20.1% |
| MASSIVE scenario Arabic | 57.0 | 19th of 26 | 89.5 | 12.9% |
| MASSIVE intent Arabic | 76.5 | 16th of 26 | 91.5 | 22.5% |
| OSACT4 offensive Arabic | 88.5 | 10th of 26 | 95.5 | 6.8% |
| Tweet sentiment Arabic | 63.0 | 14th of 26 | 80.0 | 12.9% |
| MIRACL rerank Arabic | 79.5 | 16th of 26 | 90.5 | 13.0% |
| MIRACL pairwise Arabic | 86.5 | 12th of 26 | 91.0 | 4.3% |
jaredpalmer/kev-4bRanks count the models charted on the leaderboard. Every model gets the same cases and the same request, under the benchmark rules.