system-one-bench

~/system-one-bench / compare/fastino-glide-vs-inception-mercury-decide

Head to head

GLiDE vs Mercury Decide

GLiDE scores 84.1 and Mercury Decide 82.4 on the System One Index. GLiDE is more accurate on 13 of 16 text datasets, Mercury Decide on 2, and 1 are tied. Same cases, same request, same scorer.

Overall

GLiDEMercury Decide
System One Index84.182.4
English83.083.7
Arabic84.880.8
Vision Index82.579.4
Calibration error3.6%13.1%
Median latency–276 ms
Cost per 1,000 calls$0.017Free

Text datasets, accuracy

GLiDEMercury Decide
AG News English92.092.0
SST-2 English96.095.5
SST-5 English61.552.0
DAIR Emotion English61.591.5
Banking77 English91.586.0
BoolQ English92.090.5
XNLI English English94.791.3
Typed decisions English75.071.1
XNLI French French88.082.7
XNLI Arabic Arabic83.376.0
MASSIVE scenario Arabic73.572.0
MASSIVE intent Arabic91.085.0
OSACT4 offensive Arabic95.092.0
Tweet sentiment Arabic73.562.5
MIRACL rerank Arabic87.088.0
MIRACL pairwise Arabic90.590.0

Image datasets, accuracy

GLiDEMercury Decide
SST-2, as an image94.093.0
Arabic sentiment, as an image51.539.0
Oxford-IIIT Pet84.584.5
VQAv2 yes/no89.087.0
CountBench93.593.5

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