Tsinghua ByteDance: CUDA Agent Beats Human Experts

Tsinghua University and ByteDance released CUDA Agent, an agentic reinforcement learning system that writes production CUDA kernels faster than human specialists. The model posted 100%, 100% and 92% speedups versus PyTorch compilers across KernelBench Level-1, Level-2 and Level-3 splits while beating Claude Opus 4.5 and Gemini 3 Pro by roughly 40% on the hardest tier. NVIDIA built its moat on elite hardware plus proprietary CUDA lock-in; an open model that discovers and optimizes kernels at scale through closed-loop verification now compresses that advantage, letting existing GPU fleets deliver more throughput without added silicon and underscoring Chinese teams' edge in squeezing maximum performance from constrained resources.