system-one-bench

~/system-one-bench / compare/fastino-glide-vs-jared-palmer-kev-4b

Head to head

GLiDE vs Kev 4B

GLiDE scores 84.1 and Kev 4B 76.8 on the System One Index. GLiDE is more accurate on 16 of 16 text datasets, Kev 4B on 0. Same cases, same request, same scorer.

Overall

GLiDEKev 4B
System One Index84.176.8
English83.078.3
Arabic84.874.8
Vision Index82.5–
Calibration error3.6%5.1%
Median latency–552 ms
Cost per 1,000 calls$0.017$0.010

Text datasets, accuracy

GLiDEKev 4B
AG News English92.090.5
SST-2 English96.092.5
SST-5 English61.552.0
DAIR Emotion English61.554.5
Banking77 English91.588.0
BoolQ English92.091.5
XNLI English English94.790.7
Typed decisions English75.066.3
XNLI French French88.079.3
XNLI Arabic Arabic83.372.7
MASSIVE scenario Arabic73.557.0
MASSIVE intent Arabic91.076.5
OSACT4 offensive Arabic95.088.5
Tweet sentiment Arabic73.563.0
MIRACL rerank Arabic87.079.5
MIRACL pairwise Arabic90.586.5

Image datasets, accuracy

GLiDEKev 4B
SST-2, as an image94.0–
Arabic sentiment, as an image51.5–
Oxford-IIIT Pet84.5–
VQAv2 yes/no89.0–
CountBench93.5–

Model pages: GLiDE · Kev 4B. Latency for API models is the median per call through OpenRouter, timed from a cloud server with 8 calls in flight.