Nace AI · OpenRouter API
Drex Max by Nace AI scores 84.1 on the System One Index (2nd of 30). Calibration error 6.2%, median latency 277 ms.
| Dataset | Accuracy | Rank | Best on the board | Calibration error |
|---|---|---|---|---|
| AG News English | 90.5 | 7th of 30 | 92.5 | 3.9% |
| SST-2 English | 96.0 | 3rd of 30 | 97.0 | 2.9% |
| SST-5 English | 59.0 | 7th of 30 | 62.5 | 9.6% |
| DAIR Emotion English | 65.0 | 5th of 30 | 91.5 | 8.2% |
| Banking77 English | 90.5 | 6th of 30 | 98.5 | 4.4% |
| BoolQ English | 93.0 | 1st of 30 | 93.0 | 1.9% |
| XNLI English English | 95.3 | 1st of 30 | 95.3 | 3.1% |
| Typed decisions English | 73.0 | 6th of 30 | 79.0 | 12.0% |
| XNLI French French | 91.3 | 1st of 30 | 91.3 | 3.5% |
| XNLI Arabic Arabic | 89.3 | 2nd of 32 | 90.0 | 3.8% |
| MASSIVE scenario Arabic | 73.0 | 7th of 32 | 89.5 | 11.8% |
| MASSIVE intent Arabic | 88.0 | 7th of 32 | 91.5 | 5.1% |
| OSACT4 offensive Arabic | 93.5 | 6th of 32 | 97.0 | 2.7% |
| Tweet sentiment Arabic | 69.0 | 12th of 32 | 80.0 | 4.5% |
| MIRACL rerank Arabic | 89.5 | 3rd of 32 | 90.5 | 4.5% |
| MIRACL pairwise Arabic | 89.5 | 11th of 32 | 91.5 | 3.3% |
nace-ai/drex-maxNace AI's largest decision model, the successor to Drex v2.0. Text only.
Ranks count the models charted on the leaderboard. Every model gets the same cases and the same request, under the benchmark rules.