system-one-bench

~/system-one-bench / compare/baseten-mammoth-1-vs-celeris-1-decision

Head to head

Mammoth 1 vs Celeris-1 Decision

Mammoth 1 scores 82.9 and Celeris-1 Decision 83.6 on the System One Index. Mammoth 1 is more accurate on 7 of 16 text datasets, Celeris-1 Decision on 7, and 2 are tied. Same cases, same request, same scorer.

Overall

Mammoth 1Celeris-1 Decision
System One Index82.983.6
English82.583.2
Arabic82.883.1
Vision Index––
Calibration error4.9%2.6%
Median latency––
Cost per 1,000 calls$0.037$0.014

Text datasets, accuracy

Mammoth 1Celeris-1 Decision
AG News English92.090.5
SST-2 English96.095.5
SST-5 English59.061.0
DAIR Emotion English59.072.5
Banking77 English94.589.0
BoolQ English92.089.5
XNLI English English91.391.3
Typed decisions English76.476.4
XNLI French French86.090.0
XNLI Arabic Arabic78.780.0
MASSIVE scenario Arabic75.073.5
MASSIVE intent Arabic91.088.0
OSACT4 offensive Arabic91.594.5
Tweet sentiment Arabic64.569.0
MIRACL rerank Arabic89.086.5
MIRACL pairwise Arabic90.090.0

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