Unverified50% confidenceFactExact time
The Grace Blackwell platform is currently supported primarily through CUDA 12.x and the corresponding cuDNN and NCCL versions, with PyTorch and JAX having optimized kernels for this architecture.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/14/2026
First Seen
Valid until: 12/13/2026
Sources
Related Entities
Related Claims
UnverifiedCUDA provides developers with a common computing platform that can be continuously upgraded and optimized throughout the entire lifecycle of three generations of architectures: Ampere, Hopper, and Blackwell.72% similarUnverifiedsm_103是NVIDIA Blackwell架构的CUDA计算能力标识符72% similarVerifiedCUDA is a parallel computing platform launched by NVIDIA in 2006 that has accumulated thousands of optimized libraries such as cuDNN, cuBLAS, and NCCL, forming an ecosystem moat against domestic chips.68% similarUnverifiedSince CUDA was introduced by NVIDIA in 2007, it has formed a complete accelerated library ecosystem including cuBLAS, cuDNN, and TensorRT.67% similarUnverifiedPyTorch 通过 ATen 张量库将高层 API 转译为优化的 CUDA kernel，并整合 cuDNN、cuBLAS 等 NVIDIA 优化库66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/920067API
curl https://kongchang.com/api/v1/knowledge/claims/920067MCP
get_claim(id=920067)