Skip to content

Benchmark

SWE-rebench 2025-03

Coding benchmarks compare models on programming tasks. Scores are percentages.

Models measured
14
Scores recorded
17
Sources
1
IQ weight
1.43

Score distribution

16.3%
Mean
13.9%
Median
8.4%
Std dev
7.1%
10th percentile
29.7%
90th percentile
3.9%
Min
3.9%30%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.43 for chat models, measured over 14 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard