ByteDance & Tsinghua Release CUDA Agent: Agentic RL for High-Performance Kernels
pmttyji · reddit · 2026-08-18
ByteDance Seed and Tsinghua SIA released CUDA Agent, a large-scale agentic reinforcement learning system for high-performance CUDA kernel generation. To address the specialized task of GPU optimization, the system uses a scalable data synthesis pipeline, a skill-augmented dev environment, and stable RL techniques to learn by executing code on real hardware. On KernelBench, CUDA Agent achieved speedups of 100%, 100%, and 92% over torch.compile on Level-1, Level-2, and Level-3 splits, outperforming top proprietary models like Claude Opus 4.5 and Gemini 3 Pro by about 40% on the hardest Level-3 setting.
More from Infra
- First AA-AgentPerf results: NVIDIA Vera Rubin NVL72 delivers up to 5.7x throughput per MW vs GB300 — ArtificialAnlys · 2026-10-16
- Vera Rubin NVL72 first benchmarks: up to 5.7x throughput per MW vs GB300 — ArtificialAnlys · 2026-10-16
- NVIDIA Vera Rubin NVL72 hits up to 5.7x GB300 throughput per MW on agent benchmark — ArtificialAnlys · 2026-10-16
- 14 open-source LLM observability tools to know in 2026, from Langfuse to Laminar — TheTuringPost · 2026-10-16
- Local AI is a routing problem, not a privacy feature: where each task should run — bigdata · 2026-10-16
- Moving agent work from GPUs to CPUs for durable, predictable actions — cramforce · 2026-10-16