system-one-bench

~/system-one-bench / compare/fastino-glide-vs-jericho-siahaya-naluri-279m

Head to head

GLiDE vs Naluri-279M

GLiDE scores 84.1 and Naluri-279M 50.4 on the System One Index. GLiDE is more accurate on 15 of 16 text datasets, Naluri-279M on 1. Same cases, same request, same scorer.

Overall

GLiDENaluri-279M
System One Index84.150.4
English83.052.6
Arabic84.850.7
Vision Index82.5–
Calibration error3.6%8.3%
Median latency––
Cost per 1,000 calls$0.017–

Text datasets, accuracy

GLiDENaluri-279M
AG News English92.077.5
SST-2 English96.077.0
SST-5 English61.534.0
DAIR Emotion English61.568.5
Banking77 English91.531.0
BoolQ English92.054.5
XNLI English English94.730.7
Typed decisions English75.047.3
XNLI French French88.031.3
XNLI Arabic Arabic83.330.7
MASSIVE scenario Arabic73.558.0
MASSIVE intent Arabic91.052.0
OSACT4 offensive Arabic95.082.5
Tweet sentiment Arabic73.553.0
MIRACL rerank Arabic87.028.0
MIRACL pairwise Arabic90.550.5

Image datasets, accuracy

GLiDENaluri-279M
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 · Naluri-279M. Latency for API models is the median per call through OpenRouter, timed from a cloud server with 8 calls in flight.