Benchmark Radar™
Star —

Atrex-Bench

Explore score history ↗

Evaluates coding agents on generating GPU kernels from PyTorch references across 30 operators and 440 shapes derived from production inference traces. Scoring uses a three-stage evaluator measuring compile success, numerical correctness, and speed-of-light (SOL) efficiency against a cached roofline. The benchmark includes 4 DSL backends and supports multi-vendor GPUs.

Released
2026-07-16
Institutions
Alibaba Group (China)
Evaluates
Agents, Language & Knowledge, Software & AI Compute
Openness
unknown
Importer
Claire Radar
Review status
unreviewed

Atrex-Bench paper and code

No reported scores are on record for this benchmark yet.

Which sources cite Atrex-Bench?

Related Agents, Language & Knowledge, Software & AI Compute benchmarks