~/system-one-bench / compare/google-gemini-decision-1-vs-nace-ai-drex-max
Head to head
Gemini Decision 1 scores 81.6 and Drex Max 84.1 on the System One Index. Gemini Decision 1 is more accurate on 5 of 16 text datasets, Drex Max on 9, and 2 are tied. Same cases, same request, same scorer.
| Gemini Decision 1 | Drex Max | |
|---|---|---|
| System One Index | 81.6 | 84.1 |
| English | 79.2 | 82.8 |
| Arabic | 84.5 | 84.5 |
| Vision Index | 89.2 | – |
| Calibration error | 3.6% | 6.2% |
| Median latency | – | 277 ms |
| Cost per 1,000 calls | $0.036 | $0.021 |
| Gemini Decision 1 | Drex Max | |
|---|---|---|
| AG News English | 88.5 | 90.5 |
| SST-2 English | 96.5 | 96.0 |
| SST-5 English | 59.0 | 59.0 |
| DAIR Emotion English | 59.5 | 65.0 |
| Banking77 English | 79.5 | 90.5 |
| BoolQ English | 89.0 | 93.0 |
| XNLI English English | 87.3 | 95.3 |
| Typed decisions English | 74.0 | 73.0 |
| XNLI French French | 81.3 | 91.3 |
| XNLI Arabic Arabic | 78.0 | 89.3 |
| MASSIVE scenario Arabic | 72.5 | 73.0 |
| MASSIVE intent Arabic | 87.0 | 88.0 |
| OSACT4 offensive Arabic | 95.5 | 93.5 |
| Tweet sentiment Arabic | 78.0 | 69.0 |
| MIRACL rerank Arabic | 89.5 | 89.5 |
| MIRACL pairwise Arabic | 91.0 | 89.5 |
| Gemini Decision 1 | Drex Max | |
|---|---|---|
| SST-2, as an image | 94.5 | – |
| Arabic sentiment, as an image | 75.0 | – |
| Oxford-IIIT Pet | 98.0 | – |
| VQAv2 yes/no | 86.0 | – |
| CountBench | 92.5 | – |
Model pages: Gemini Decision 1 · Drex Max. Latency for API models is the median per call through OpenRouter, timed from a cloud server with 8 calls in flight.