system-one-bench

~/system-one-bench / compare/baseten-joey-1-vs-scaledown-classify-1

Head to head

Joey 1 vs Classify-1

Joey 1 scores 73.3 and Classify-1 30.6 on the System One Index. Joey 1 is more accurate on 15 of 16 text datasets, Classify-1 on 1. Same cases, same request, same scorer.

Overall

Joey 1Classify-1
System One Index73.330.6
English72.930.7
Arabic73.530.6
Vision Indexโ€“โ€“
Calibration error2.1%3.2%
Median latencyโ€“254 ms
Cost per 1,000 calls$0.004$0.018

Text datasets, accuracy

Joey 1Classify-1
AG News English85.026.5
SST-2 English86.050.0
SST-5 English40.513.0
DAIR Emotion English52.559.0
Banking77 English93.01.0
BoolQ English80.038.0
XNLI English English84.030.7
Typed decisions English62.427.2
XNLI French French74.730.7
XNLI Arabic Arabic64.730.7
MASSIVE scenario Arabic71.53.0
MASSIVE intent Arabic84.53.5
OSACT4 offensive Arabic83.080.0
Tweet sentiment Arabic66.035.0
MIRACL rerank Arabic62.012.0
MIRACL pairwise Arabic83.050.0

Model pages: Joey 1 ยท Classify-1. Latency for API models is the median per call through OpenRouter, timed from a cloud server with 8 calls in flight.