Skip to content

Benchmark

Vals SWE-bench

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

Models measured
83
Scores recorded
83
Sources
1
IQ weight
3.33

Score distribution

73.7%
Mean
75%
Median
14.7%
Std dev
57.6%
10th percentile
93%
90th percentile
7.8%
Min
7.8%97%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 3.33 for chat models, measured over 83 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

Showing 51–83 of 83 results.