tilelang-b3e332cd·2 events·first seen Aliases: TileLang
TileLang is an open-source domain-specific language (DSL) aimed at simplifying the development of high-performance compute kernels for GPUs, CPUs, and accelerators. The project is trending on GitHub with 6,871 total stars and 109 new stars today. It is relevant to the AI/ML inference and training infrastructure space, where custom kernel development is a key performance bottleneck.
DeepSeek has released DeepSeek-V3.2-Exp, an experimental model built on V3.1-Terminus that introduces DeepSeek Sparse Attention (DSA), a fine-grained sparse attention mechanism designed to improve long-context performance and reduce compute costs during training and inference. Benchmarks indicate V3.2-Exp performs on par with V3.1-Terminus while achieving efficiency gains. The release is accompanied by a 50%+ API price reduction effective immediately, open-weights release on Hugging Face, a technical report, and GPU kernel code in TileLang and CUDA.