Liquid AI ยท OpenRouter API
d1-ultra by Liquid AI scores 82.9 on the System One Index (3rd of 30) and 85.6 on the Vision Index (2nd of 11). Calibration error 1.0%, median latency 168 ms.
| Dataset | Accuracy | Rank | Best on the board | Calibration error |
|---|---|---|---|---|
| AG News English | 91.0 | 5th of 30 | 92.5 | 3.2% |
| SST-2 English | 95.5 | 4th of 30 | 97.0 | 2.1% |
| SST-5 English | 61.5 | 3rd of 30 | 62.5 | 8.2% |
| DAIR Emotion English | 63.0 | 8th of 30 | 91.5 | 14.2% |
| Banking77 English | 83.5 | 17th of 30 | 98.5 | 6.6% |
| BoolQ English | 91.5 | 3rd of 30 | 93.0 | 3.2% |
| XNLI English English | 92.7 | 3rd of 30 | 95.3 | 11.0% |
| Typed decisions English | 75.0 | 3rd of 30 | 79.0 | 3.1% |
| XNLI French French | 83.3 | 6th of 30 | 91.3 | 10.0% |
| XNLI Arabic Arabic | 76.0 | 7th of 32 | 90.0 | 9.8% |
| MASSIVE scenario Arabic | 76.0 | 6th of 32 | 89.5 | 6.9% |
| MASSIVE intent Arabic | 87.0 | 8th of 32 | 91.5 | 6.2% |
| OSACT4 offensive Arabic | 94.5 | 4th of 32 | 97.0 | 5.4% |
| Tweet sentiment Arabic | 76.0 | 2nd of 32 | 80.0 | 8.1% |
| MIRACL rerank Arabic | 89.0 | 4th of 32 | 90.5 | 4.6% |
| MIRACL pairwise Arabic | 91.0 | 3rd of 32 | 91.5 | 2.1% |
| Dataset | Accuracy | Rank | Best on the board | Calibration error |
|---|---|---|---|---|
| SST-2, as an image | 97.0 | 1st of 11 | 97.0 | 4.3% |
| Arabic sentiment, as an image | 62.5 | 4th of 11 | 70.0 | 5.9% |
| Oxford-IIIT Pet | 87.5 | 4th of 11 | 92.5 | 6.9% |
| VQAv2 yes/no | 89.5 | 1st of 11 | 89.5 | 3.2% |
| CountBench | 91.5 | 5th of 11 | 95.5 | 7.6% |
liquid/d1-ultraLiquid AI's largest decision model. 32K context. Doesn't answer Box questions.
Ranks count the models charted on the leaderboard. Every model gets the same cases and the same request, under the benchmark rules.