Skip to content

Benchmark

SciCode

Coding benchmarks compare models on programming tasks. Scores are percentages. Higher is better.

Models measured
154
Scores recorded
265
Sources
3
IQ weight
1.71

Score distribution

47.7%
Mean
50.1%
Median
10.7%
Std dev
35.4%
10th percentile
58.8%
90th percentile
1.3%
Min
1.3%91.8%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.71 for chat models, measured over 154 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

Showing 251–265 of 265 results.

Data credits: Data by LLM Stats · Epoch AI Benchmarking Hub, CC BY 4.0 (licence) · BenchBoard by SYuan03 (github.com/SYuan03/benchboard). MIT covers its code and data organization; no results licence is inferred. Numeric relays retain the original stating-source links.