shipfeedAI news, curated daily

04:10:16 CET
24 JUN04:10:16shipfeed
pull to refreshlast sync
Just in — 30 new
§ evals · storyline

ParallelKernelBench: Frontier LLMs can't write fast multi-GPU kernels (yet)

ParallelKernelBench benchmark shows frontier LLMs solve fewer than a third of multi-GPU CUDA kernel tasks despite occasionally outperforming public implementations.

yesterday · · primary fetch1 sourceupdated yesterday ·

ParallelKernelBench tests whether LLMs can write fast multi-GPU CUDA kernels across 87 real workloads.

The best model solves under a third, but a few generated kernels beat any public implementation.

read full article on together.ai
§ sources1 publication · timeline below
  1. together.aiParallelKernelBench: Frontier LLMs can't write fast multi-GPU kernels (yet)primary