Head to head
Drex v2.0 scores 78.6 and GPT-6 Luna Decisions 76.0 on the System One Index. Drex v2.0 is more accurate on 10 of 16 text datasets, GPT-6 Luna Decisions on 6. Same cases, same request, same scorer.
| Drex v2.0 | GPT-6 Luna Decisions | |
|---|---|---|
| System One Index | 78.6 | 76.0 |
| English | 77.8 | 75.2 |
| Arabic | 79.6 | 78.8 |
| Vision Index | – | 81.6 |
| Calibration error | 1.5% | 9.8% |
| Median latency | 294 ms | 215 ms |
| Cost per 1,000 calls | $0.015 | $0.040 |
| Drex v2.0 | GPT-6 Luna Decisions | |
|---|---|---|
| AG News English | 88.0 | 87.5 |
| SST-2 English | 94.0 | 93.5 |
| SST-5 English | 56.0 | 61.5 |
| DAIR Emotion English | 55.5 | 58.5 |
| Banking77 English | 80.5 | 77.5 |
| BoolQ English | 92.5 | 85.5 |
| XNLI English English | 87.3 | 66.7 |
| Typed decisions English | 68.7 | 71.0 |
| XNLI French French | 77.3 | 62.0 |
| XNLI Arabic Arabic | 71.3 | 54.0 |
| MASSIVE scenario Arabic | 63.5 | 66.0 |
| MASSIVE intent Arabic | 82.5 | 84.5 |
| OSACT4 offensive Arabic | 94.5 | 93.5 |
| Tweet sentiment Arabic | 69.5 | 80.0 |
| MIRACL rerank Arabic | 87.0 | 86.0 |
| MIRACL pairwise Arabic | 89.0 | 87.5 |
| Drex v2.0 | GPT-6 Luna Decisions | |
|---|---|---|
| SST-2, as an image | – | 88.5 |
| Arabic sentiment, as an image | – | 67.0 |
| Oxford-IIIT Pet | – | 87.0 |
| VQAv2 yes/no | – | 80.5 |
| CountBench | – | 85.0 |
Model pages: Drex v2.0 · GPT-6 Luna Decisions. Latency for API models is the median per call through OpenRouter, timed from a cloud server with 8 calls in flight.