Perplexity ยท OpenRouter API
Decider V1.2 by Perplexity scores 84.6 on the System One Index (1st of 30) and 83.8 on the Vision Index (5th of 11). Calibration error 0.7%, median latency 178 ms.
| Dataset | Accuracy | Rank | Best on the board | Calibration error |
|---|---|---|---|---|
| AG News English | 91.5 | 3rd of 30 | 92.5 | 4.9% |
| SST-2 English | 95.5 | 4th of 30 | 97.0 | 2.6% |
| SST-5 English | 60.0 | 5th of 30 | 62.5 | 7.0% |
| DAIR Emotion English | 60.0 | 9th of 30 | 91.5 | 8.5% |
| Banking77 English | 88.0 | 10th of 30 | 98.5 | 3.7% |
| BoolQ English | 91.5 | 3rd of 30 | 93.0 | 2.4% |
| XNLI English English | 94.0 | 2nd of 30 | 95.3 | 4.2% |
| Typed decisions English | 75.1 | 2nd of 30 | 79.0 | 1.6% |
| XNLI French French | 90.0 | 2nd of 30 | 91.3 | 5.6% |
| XNLI Arabic Arabic | 90.0 | 1st of 32 | 90.0 | 7.7% |
| MASSIVE scenario Arabic | 79.0 | 5th of 32 | 89.5 | 5.3% |
| MASSIVE intent Arabic | 91.0 | 2nd of 32 | 91.5 | 4.9% |
| OSACT4 offensive Arabic | 95.0 | 3rd of 32 | 97.0 | 4.4% |
| Tweet sentiment Arabic | 75.5 | 6th of 32 | 80.0 | 3.7% |
| MIRACL rerank Arabic | 86.5 | 13th of 32 | 90.5 | 4.7% |
| MIRACL pairwise Arabic | 91.5 | 1st of 32 | 91.5 | 5.5% |
| Dataset | Accuracy | Rank | Best on the board | Calibration error |
|---|---|---|---|---|
| SST-2, as an image | 94.5 | 4th of 11 | 97.0 | 4.9% |
| Arabic sentiment, as an image | 57.0 | 6th of 11 | 70.0 | 5.9% |
| Oxford-IIIT Pet | 85.5 | 6th of 11 | 92.5 | 17.8% |
| VQAv2 yes/no | 86.5 | 9th of 11 | 89.5 | 3.1% |
| CountBench | 95.5 | 1st of 11 | 95.5 | 4.4% |
perplexity/pplx-decider-v1.227B. Replaces Decider V1.1 27B. Doesn't answer Box questions.
Ranks count the models charted on the leaderboard. Every model gets the same cases and the same request, under the benchmark rules.