~/system-one-bench / compare/meragpt-decider-1-vs-perplexity-decider-v1-2
Head to head
Decider 1 scores 61.8 and Decider V1.2 84.6 on the System One Index. Decider 1 is more accurate on 1 of 16 text datasets, Decider V1.2 on 15. Same cases, same request, same scorer.
| Decider 1 | Decider V1.2 | |
|---|---|---|
| System One Index | 61.8 | 84.6 |
| English | 67.4 | 82.0 |
| Arabic | 54.7 | 86.9 |
| Vision Index | – | 83.8 |
| Calibration error | 7.9% | 0.7% |
| Median latency | – | 178 ms |
| Cost per 1,000 calls | $0.012 | $0.007 |
| Decider 1 | Decider V1.2 | |
|---|---|---|
| AG News English | 87.5 | 91.5 |
| SST-2 English | 93.0 | 95.5 |
| SST-5 English | 55.5 | 60.0 |
| DAIR Emotion English | 67.5 | 60.0 |
| Banking77 English | 1.0 | 88.0 |
| BoolQ English | 80.0 | 91.5 |
| XNLI English English | 80.7 | 94.0 |
| Typed decisions English | 74.2 | 75.1 |
| XNLI French French | 67.3 | 90.0 |
| XNLI Arabic Arabic | 67.3 | 90.0 |
| MASSIVE scenario Arabic | 3.0 | 79.0 |
| MASSIVE intent Arabic | 3.5 | 91.0 |
| OSACT4 offensive Arabic | 86.0 | 95.0 |
| Tweet sentiment Arabic | 64.0 | 75.5 |
| MIRACL rerank Arabic | 81.0 | 86.5 |
| MIRACL pairwise Arabic | 78.0 | 91.5 |
| Decider 1 | Decider V1.2 | |
|---|---|---|
| SST-2, as an image | – | 94.5 |
| Arabic sentiment, as an image | – | 57.0 |
| Oxford-IIIT Pet | – | 85.5 |
| VQAv2 yes/no | – | 86.5 |
| CountBench | – | 95.5 |
Model pages: Decider 1 · Decider V1.2. Latency for API models is the median per call through OpenRouter, timed from a cloud server with 8 calls in flight.