Liquid AI · OpenRouter API
d1 by Liquid AI scores 82.7 on the System One Index (2nd of 24). Calibration error 2.9%, median latency 188 ms.
| Dataset | Accuracy | Rank | Best on the board | Calibration error |
|---|---|---|---|---|
| AG News English | 90.0 | 8th of 24 | 92.5 | 6.1% |
| SST-2 English | 95.5 | 4th of 24 | 97.0 | 2.0% |
| SST-5 English | 56.5 | 7th of 24 | 62.5 | 8.0% |
| DAIR Emotion English | 63.5 | 5th of 24 | 91.5 | 11.9% |
| Banking77 English | 89.0 | 7th of 24 | 98.5 | 3.6% |
| BoolQ English | 90.0 | 6th of 24 | 92.5 | 4.2% |
| XNLI English English | 91.3 | 2nd of 24 | 93.3 | 2.5% |
| Typed decisions English | 74.5 | 2nd of 24 | 79.0 | 3.2% |
| XNLI French French | 84.7 | 3rd of 24 | 90.7 | 7.7% |
| XNLI Arabic Arabic | 80.0 | 3rd of 26 | 89.3 | 6.1% |
| MASSIVE scenario Arabic | 73.0 | 5th of 26 | 89.5 | 8.2% |
| MASSIVE intent Arabic | 90.0 | 3rd of 26 | 91.5 | 4.1% |
| OSACT4 offensive Arabic | 95.5 | 1st of 26 | 95.5 | 5.1% |
| Tweet sentiment Arabic | 67.5 | 9th of 26 | 80.0 | 12.5% |
| MIRACL rerank Arabic | 90.5 | 1st of 26 | 90.5 | 3.0% |
| MIRACL pairwise Arabic | 91.0 | 1st of 26 | 91.0 | 3.5% |
liquid/d1Ranks count the models charted on the leaderboard. Every model gets the same cases and the same request, under the benchmark rules.