system-one-bench

~/system-one-bench / compare/baseten-joey-1-vs-patrick-18-dexy-decision-9b

Head to head

Joey 1 vs Dexy-Decision-9B

Joey 1 scores 73.3 and Dexy-Decision-9B 84.7 on the System One Index. Joey 1 is more accurate on 0 of 16 text datasets, Dexy-Decision-9B on 15, and 1 are tied. Same cases, same request, same scorer.

Overall

Joey 1Dexy-Decision-9B
System One Index73.384.7
English72.982.3
Arabic73.586.8
Vision Index––
Calibration error2.1%3.5%
Median latency––
Cost per 1,000 calls$0.004–

Text datasets, accuracy

Joey 1Dexy-Decision-9B
AG News English85.091.5
SST-2 English86.095.5
SST-5 English40.560.0
DAIR Emotion English52.554.0
Banking77 English93.093.0
BoolQ English80.089.5
XNLI English English84.094.7
Typed decisions English62.480.0
XNLI French French74.789.3
XNLI Arabic Arabic64.789.3
MASSIVE scenario Arabic71.589.5
MASSIVE intent Arabic84.591.5
OSACT4 offensive Arabic83.091.5
Tweet sentiment Arabic66.067.0
MIRACL rerank Arabic62.088.0
MIRACL pairwise Arabic83.091.0

Model pages: Joey 1 · Dexy-Decision-9B. Latency for API models is the median per call through OpenRouter, timed from a cloud server with 8 calls in flight.