system-one-bench

~/system-one-bench / compare/baseten-joey-1-vs-together-ai-tev1-4b-experimental

Head to head

Joey 1 vs Tev1 4B Experimental

Joey 1 scores 73.3 and Tev1 4B Experimental 71.5 on the System One Index. Joey 1 is more accurate on 3 of 16 text datasets, Tev1 4B Experimental on 13. Same cases, same request, same scorer.

Overall

Joey 1Tev1 4B Experimental
System One Index73.371.5
English72.966.4
Arabic73.575.8
Vision Indexโ€“โ€“
Calibration error2.1%5.3%
Median latencyโ€“232 ms
Cost per 1,000 calls$0.004$0.019

Text datasets, accuracy

Joey 1Tev1 4B Experimental
AG News English85.091.5
SST-2 English86.094.0
SST-5 English40.543.0
DAIR Emotion English52.556.5
Banking77 English93.01.0
BoolQ English80.091.0
XNLI English English84.087.3
Typed decisions English62.467.0
XNLI French French74.782.7
XNLI Arabic Arabic64.776.0
MASSIVE scenario Arabic71.556.5
MASSIVE intent Arabic84.570.0
OSACT4 offensive Arabic83.084.5
Tweet sentiment Arabic66.070.5
MIRACL rerank Arabic62.085.5
MIRACL pairwise Arabic83.087.5

Model pages: Joey 1 ยท Tev1 4B Experimental. Latency for API models is the median per call through OpenRouter, timed from a cloud server with 8 calls in flight.