Skip to content

Benchmark

SWE-rebench 2025-07

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

Models measured
25
Scores recorded
31
Sources
1
IQ weight
1.76

Score distribution

13.6%
Mean
14.7%
Median
5.8%
Std dev
6.7%
10th percentile
22%
90th percentile
0.667%
Min
0.667%26.7%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.76 for chat models, measured over 25 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard