TypeSafe ยท OpenRouter API
Jev 1.13 by TypeSafe scores 79.0 on the System One Index (9th of 24) and 83.9 on the Vision Index (2nd of 8). Calibration error 6.1%, median latency 150 ms.
| Dataset | Accuracy | Rank | Best on the board | Calibration error |
|---|---|---|---|---|
| AG News English | 88.5 | 14th of 24 | 92.5 | 8.6% |
| SST-2 English | 94.5 | 10th of 24 | 97.0 | 6.4% |
| SST-5 English | 62.0 | 2nd of 24 | 62.5 | 16.1% |
| DAIR Emotion English | 56.0 | 17th of 24 | 91.5 | 30.4% |
| Banking77 English | 79.0 | 16th of 24 | 98.5 | 10.1% |
| BoolQ English | 89.5 | 9th of 24 | 92.5 | 3.9% |
| XNLI English English | 86.0 | 12th of 24 | 93.3 | 8.2% |
| Typed decisions English | 72.7 | 5th of 24 | 79.0 | 4.0% |
| XNLI French French | 76.7 | 14th of 24 | 90.7 | 12.2% |
| XNLI Arabic Arabic | 69.3 | 16th of 26 | 89.3 | 14.0% |
| MASSIVE scenario Arabic | 63.0 | 13th of 26 | 89.5 | 21.0% |
| MASSIVE intent Arabic | 84.0 | 10th of 26 | 91.5 | 3.5% |
| OSACT4 offensive Arabic | 90.0 | 7th of 26 | 95.5 | 5.5% |
| Tweet sentiment Arabic | 76.0 | 2nd of 26 | 80.0 | 9.7% |
| MIRACL rerank Arabic | 86.5 | 9th of 26 | 90.5 | 6.1% |
| MIRACL pairwise Arabic | 90.5 | 3rd of 26 | 91.0 | 2.7% |
| Dataset | Accuracy | Rank | Best on the board | Calibration error |
|---|---|---|---|---|
| SST-2, as an image | 94.5 | 4th of 8 | 96.0 | 6.0% |
| Arabic sentiment, as an image | 59.0 | 3rd of 8 | 67.0 | 10.5% |
| Oxford-IIIT Pet | 83.5 | 7th of 8 | 88.5 | 5.1% |
| VQAv2 yes/no | 88.0 | 2nd of 8 | 89.5 | 3.0% |
| CountBench | 94.5 | 1st of 8 | 94.5 | 4.6% |
typesafe/jev-1.13Ranks count the models charted on the leaderboard. Every model gets the same cases and the same request, under the benchmark rules.