~/system-one-bench / compare/google-gemini-decision-1-vs-typesafe-jev-1-13
Head to head
Gemini Decision 1 scores 81.6 and Jev 1.13 79.0 on the System One Index. Gemini Decision 1 is more accurate on 13 of 16 text datasets, Jev 1.13 on 2, and 1 are tied. Same cases, same request, same scorer.
| Gemini Decision 1 | Jev 1.13 | |
|---|---|---|
| System One Index | 81.6 | 79.0 |
| English | 79.2 | 78.5 |
| Arabic | 84.5 | 79.9 |
| Vision Index | 89.2 | 83.9 |
| Calibration error | 3.6% | 6.1% |
| Median latency | – | 150 ms |
| Cost per 1,000 calls | $0.036 | $0.028 |
| Gemini Decision 1 | Jev 1.13 | |
|---|---|---|
| AG News English | 88.5 | 88.5 |
| SST-2 English | 96.5 | 94.5 |
| SST-5 English | 59.0 | 62.0 |
| DAIR Emotion English | 59.5 | 56.0 |
| Banking77 English | 79.5 | 79.0 |
| BoolQ English | 89.0 | 89.5 |
| XNLI English English | 87.3 | 86.0 |
| Typed decisions English | 74.0 | 72.7 |
| XNLI French French | 81.3 | 76.7 |
| XNLI Arabic Arabic | 78.0 | 69.3 |
| MASSIVE scenario Arabic | 72.5 | 63.0 |
| MASSIVE intent Arabic | 87.0 | 84.0 |
| OSACT4 offensive Arabic | 95.5 | 90.0 |
| Tweet sentiment Arabic | 78.0 | 76.0 |
| MIRACL rerank Arabic | 89.5 | 86.5 |
| MIRACL pairwise Arabic | 91.0 | 90.5 |
| Gemini Decision 1 | Jev 1.13 | |
|---|---|---|
| SST-2, as an image | 94.5 | 94.5 |
| Arabic sentiment, as an image | 75.0 | 59.0 |
| Oxford-IIIT Pet | 98.0 | 83.5 |
| VQAv2 yes/no | 86.0 | 88.0 |
| CountBench | 92.5 | 94.5 |
Model pages: Gemini Decision 1 · Jev 1.13. Latency for API models is the median per call through OpenRouter, timed from a cloud server with 8 calls in flight.