system-one-bench

~/system-one-bench / compare/baseten-elephant-1-vs-perplexity-decider-v1-2

Head to head

Elephant 1 vs Decider V1.2

Elephant 1 scores 82.1 and Decider V1.2 84.6 on the System One Index. Elephant 1 is more accurate on 2 of 16 text datasets, Decider V1.2 on 12, and 2 are tied. Same cases, same request, same scorer.

Overall

Elephant 1Decider V1.2
System One Index82.184.6
English81.382.0
Arabic82.086.9
Vision Index–83.8
Calibration error3.7%0.7%
Median latency–178 ms
Cost per 1,000 calls$0.013$0.007

Text datasets, accuracy

Elephant 1Decider V1.2
AG News English89.091.5
SST-2 English94.595.5
SST-5 English58.560.0
DAIR Emotion English54.560.0
Banking77 English94.588.0
BoolQ English91.091.5
XNLI English English94.094.0
Typed decisions English74.275.1
XNLI French French90.090.0
XNLI Arabic Arabic79.390.0
MASSIVE scenario Arabic69.079.0
MASSIVE intent Arabic89.591.0
OSACT4 offensive Arabic91.095.0
Tweet sentiment Arabic66.075.5
MIRACL rerank Arabic88.086.5
MIRACL pairwise Arabic91.091.5

Image datasets, accuracy

Elephant 1Decider V1.2
SST-2, as an image–94.5
Arabic sentiment, as an image–57.0
Oxford-IIIT Pet–85.5
VQAv2 yes/no–86.5
CountBench–95.5

Model pages: Elephant 1 · Decider V1.2. Latency for API models is the median per call through OpenRouter, timed from a cloud server with 8 calls in flight.