GPU Infrastructure Inference Optimization LLM Serving

Tile Kernels

Optimized GPU kernels for LLM operations, built with TileLang.

Open-source TileLang-based kernel library for optimized GPU kernels for LLM operations, including MoE routing, quantization, transpose, engram, and manifold hyperconnection kernels.

Recent stories

1 linked story

releasePRIMARY2026-04-23

DeepSeek releases Tile Kernels with Engram, mHC, and FP4/FP8 ops for SM90 and SM100 GPUs

DeepSeek published Tile Kernels, an open-source TileLang repo covering Engram, mHC, MoE routing, and FP4/FP8 kernels, with claims that some are already used in internal training and inference. That matters because it exposes reusable low-level performance work behind DeepSeek’s stack instead of keeping the kernels fully private.