Skip to content

Benchmark

AHELM: PARADE - EM

Scores are percentages.

Models measured
12
Scores recorded
60
Sources
1
IQ weight
0.38

Score distribution

46.5%
Mean
46%
Median
26.2%
Std dev
9.8%
10th percentile
83%
90th percentile
2.7%
Min
2.7%97%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 0.38 for chat models, measured over 9 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
51Gemini 2.5 Pro Preview 05-06subset: occupation, voice: female· AHELM (HELM Audio) data15.2%
52Gemini 2.5 Pro Preview 05-06subset: occupation, voice: male· AHELM (HELM Audio) data14.4%
53gemini-1.5-flash-002subset: occupation, voice: male· AHELM (HELM Audio) data11.4%
54gemini-1.5-pro-001subset: occupation, voice: male· AHELM (HELM Audio) data9.8%
55gemini-1.5-pro-001subset: occupation, voice: female· AHELM (HELM Audio) data8.9%
56gemini-1.5-flash-002subset: occupation, voice: female· AHELM (HELM Audio) data8.4%
57gemini-1.5-pro-002subset: occupation, voice: female· AHELM (HELM Audio) data5.1%
58gemini-1.5-pro-002subset: occupation, voice: male· AHELM (HELM Audio) data3.8%
59gemini-1.5-flash-001subset: occupation, voice: male· AHELM (HELM Audio) data3%
60gemini-1.5-flash-001subset: occupation, voice: female· AHELM (HELM Audio) data2.7%

Showing 51–60 of 60 results.