ParallelKernelBench：前沿 LLM 尚无法编写快速多 GPU 内核
原标题：ParallelKernelBench: Frontier LLMs can't write fast multi-GPU kernels (yet)
AI 摘要
Together AI 发布了 ParallelKernelBench（PKB），一个用于评估多 GPU 内核生成的基准测试框架，包含 87 个来自真实代码库的问题，要求模型将 PyTorch + NCCL 实现替换为直接通过 NVLink 通信的 CUDA 内核。测试前沿模型如 GPT-5.5、Gemini 3 Pro 和 Opus 4.7 后发现，正确率不足三分之一，且只有少数解决方案优于基线。模型在协调、数据分区和通信机制选择上存在困难，但代理式反馈循环能带来适度改进。
正文节选
LLMs have gotten surprisingly good at writing GPU kernels[1][2][3], but almost all current benchmarks measuring that progress are single-GPU. In production, communication is often the bottleneck: communication overhead can account for over 20% of inference latency[4], and that gap keeps widening as compute scales faster than interconnect bandwidth. ParallelKernelBench (PKB) offers a benchmark and evaluation framework for multi-GPU kernel generation and includes 87 problems from real codebases wh