system-one-bench

~/system-one-bench / compare/celeris-1-decision-vs-nace-ai-drex-v1-5

Head to head

Celeris-1 Decision vs Drex v1.5

Celeris-1 Decision scores 83.6 and Drex v1.5 77.6 on the System One Index. Celeris-1 Decision is more accurate on 15 of 16 text datasets, Drex v1.5 on 1. Same cases, same request, same scorer.

Overall

Celeris-1 DecisionDrex v1.5
System One Index83.677.6
English83.277.6
Arabic83.177.0
Vision Indexโ€“โ€“
Calibration error2.6%10.8%
Median latencyโ€“198 ms
Cost per 1,000 calls$0.014$0.010

Text datasets, accuracy

Celeris-1 DecisionDrex v1.5
AG News English90.590.0
SST-2 English95.594.5
SST-5 English61.053.5
DAIR Emotion English72.553.5
Banking77 English89.087.0
BoolQ English89.589.0
XNLI English English91.388.7
Typed decisions English76.464.8
XNLI French French90.082.0
XNLI Arabic Arabic80.074.0
MASSIVE scenario Arabic73.558.5
MASSIVE intent Arabic88.077.5
OSACT4 offensive Arabic94.590.5
Tweet sentiment Arabic69.063.5
MIRACL rerank Arabic86.589.0
MIRACL pairwise Arabic90.086.0

Model pages: Celeris-1 Decision ยท Drex v1.5. Latency for API models is the median per call through OpenRouter, timed from a cloud server with 8 calls in flight.