Atlas

Benchmarks

← All benchmarks

AlgoTune

Code · 2025-07-19

AlgoTune evaluates models on optimizing algorithmic Python code for speed while preserving correctness. Scores are reported as speedup factors, so higher values indicate stronger optimization performance.

Top models (higher is better)

ModelScore
GPT-5.22.0
Gemini 3.1 Pro Preview2.0
GPT-5.41.9
Gemini 3 Pro Preview1.8
Opus 4.51.8
O4 Mini1.7
GPT-51.7
Sonnet 4.51.5
GLM-4.51.5
Gemini 2.5 Pro1.5
Opus 4.61.5
gpt-oss-120b1.4
GPT-5 Mini1.4
Opus 4.11.3
Opus 41.3
Loading Atlas data…