~/system-one-bench / compare/fastino-glide-vs-scaledown-classify-1
Head to head
GLiDE scores 84.1 and Classify-1 30.6 on the System One Index. GLiDE is more accurate on 16 of 16 text datasets, Classify-1 on 0. Same cases, same request, same scorer.
| GLiDE | Classify-1 | |
|---|---|---|
| System One Index | 84.1 | 30.6 |
| English | 83.0 | 30.7 |
| Arabic | 84.8 | 30.6 |
| Vision Index | 82.5 | – |
| Calibration error | 3.6% | 3.2% |
| Median latency | – | 254 ms |
| Cost per 1,000 calls | $0.017 | $0.018 |
| GLiDE | Classify-1 | |
|---|---|---|
| AG News English | 92.0 | 26.5 |
| SST-2 English | 96.0 | 50.0 |
| SST-5 English | 61.5 | 13.0 |
| DAIR Emotion English | 61.5 | 59.0 |
| Banking77 English | 91.5 | 1.0 |
| BoolQ English | 92.0 | 38.0 |
| XNLI English English | 94.7 | 30.7 |
| Typed decisions English | 75.0 | 27.2 |
| XNLI French French | 88.0 | 30.7 |
| XNLI Arabic Arabic | 83.3 | 30.7 |
| MASSIVE scenario Arabic | 73.5 | 3.0 |
| MASSIVE intent Arabic | 91.0 | 3.5 |
| OSACT4 offensive Arabic | 95.0 | 80.0 |
| Tweet sentiment Arabic | 73.5 | 35.0 |
| MIRACL rerank Arabic | 87.0 | 12.0 |
| MIRACL pairwise Arabic | 90.5 | 50.0 |
| GLiDE | Classify-1 | |
|---|---|---|
| SST-2, as an image | 94.0 | – |
| Arabic sentiment, as an image | 51.5 | – |
| Oxford-IIIT Pet | 84.5 | – |
| VQAv2 yes/no | 89.0 | – |
| CountBench | 93.5 | – |
Model pages: GLiDE · Classify-1. Latency for API models is the median per call through OpenRouter, timed from a cloud server with 8 calls in flight.