system-one-bench

~/system-one-bench / compare/baseten-elephant-1-vs-fastino-glide

Head to head

Elephant 1 vs GLiDE

Elephant 1 scores 82.1 and GLiDE 84.1 on the System One Index. Elephant 1 is more accurate on 4 of 16 text datasets, GLiDE on 12. Same cases, same request, same scorer.

Overall

Elephant 1GLiDE
System One Index82.184.1
English81.383.0
Arabic82.084.8
Vision Index–82.5
Calibration error3.7%3.6%
Median latency––
Cost per 1,000 calls$0.013$0.017

Text datasets, accuracy

Elephant 1GLiDE
AG News English89.092.0
SST-2 English94.596.0
SST-5 English58.561.5
DAIR Emotion English54.561.5
Banking77 English94.591.5
BoolQ English91.092.0
XNLI English English94.094.7
Typed decisions English74.275.0
XNLI French French90.088.0
XNLI Arabic Arabic79.383.3
MASSIVE scenario Arabic69.073.5
MASSIVE intent Arabic89.591.0
OSACT4 offensive Arabic91.095.0
Tweet sentiment Arabic66.073.5
MIRACL rerank Arabic88.087.0
MIRACL pairwise Arabic91.090.5

Image datasets, accuracy

Elephant 1GLiDE
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: Elephant 1 · GLiDE. Latency for API models is the median per call through OpenRouter, timed from a cloud server with 8 calls in flight.