system-one-bench

~/system-one-bench / compare/baseten-mammoth-1-vs-inception-mercury-decide

Head to head

Mammoth 1 vs Mercury Decide

Mammoth 1 scores 82.9 and Mercury Decide 82.4 on the System One Index. Mammoth 1 is more accurate on 11 of 16 text datasets, Mercury Decide on 2, and 3 are tied. Same cases, same request, same scorer.

Overall

Mammoth 1Mercury Decide
System One Index82.982.4
English82.583.7
Arabic82.880.8
Vision Index–79.4
Calibration error4.9%13.1%
Median latency–276 ms
Cost per 1,000 calls$0.037Free

Text datasets, accuracy

Mammoth 1Mercury Decide
AG News English92.092.0
SST-2 English96.095.5
SST-5 English59.052.0
DAIR Emotion English59.091.5
Banking77 English94.586.0
BoolQ English92.090.5
XNLI English English91.391.3
Typed decisions English76.471.1
XNLI French French86.082.7
XNLI Arabic Arabic78.776.0
MASSIVE scenario Arabic75.072.0
MASSIVE intent Arabic91.085.0
OSACT4 offensive Arabic91.592.0
Tweet sentiment Arabic64.562.5
MIRACL rerank Arabic89.088.0
MIRACL pairwise Arabic90.090.0

Image datasets, accuracy

Mammoth 1Mercury Decide
SST-2, as an image–93.0
Arabic sentiment, as an image–39.0
Oxford-IIIT Pet–84.5
VQAv2 yes/no–87.0
CountBench–93.5

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