system-one-bench

~/system-one-bench / compare/celeris-1-decision-vs-vllm-semantic-router-d3-lite

Head to head

Celeris-1 Decision vs d3 lite

Celeris-1 Decision scores 83.6 and d3 lite 74.9 on the System One Index. Celeris-1 Decision is more accurate on 15 of 16 text datasets, d3 lite on 1. Same cases, same request, same scorer.

Overall

Celeris-1 Decisiond3 lite
System One Index83.674.9
English83.274.3
Arabic83.174.7
Vision Index––
Calibration error2.6%7.1%
Median latency––
Cost per 1,000 calls$0.014–

Text datasets, accuracy

Celeris-1 Decisiond3 lite
AG News English90.585.5
SST-2 English95.586.0
SST-5 English61.050.5
DAIR Emotion English72.557.0
Banking77 English89.091.0
BoolQ English89.581.0
XNLI English English91.390.7
Typed decisions English76.452.6
XNLI French French90.081.3
XNLI Arabic Arabic80.072.7
MASSIVE scenario Arabic73.558.0
MASSIVE intent Arabic88.077.5
OSACT4 offensive Arabic94.580.5
Tweet sentiment Arabic69.067.0
MIRACL rerank Arabic86.582.5
MIRACL pairwise Arabic90.084.5

Model pages: Celeris-1 Decision · d3 lite. Latency for API models is the median per call through OpenRouter, timed from a cloud server with 8 calls in flight.