search
The Matrix
Trends
- 1
Chinese AI lab DeepSeek has released DeepGEMM, an open-source library of clean and efficient BLAS matrix-multiplication kernels written in CUDA for NVIDIA GPUs. The project is drawing developer attention on GitHub, where it ranks among trending repositories, as efficient low-level kernels are key to fast and cheap AI model training and inference.
- 2OpenAI Preprints Claim Faster Matrix Multiplication Methods●OpenAI Preprints Break Integer and Matrix Multiplication Barriers
OpenAI has released preprints describing new algorithms that it says break longstanding barriers in integer and matrix multiplication. The work suggests faster approaches to problems central to computing and scientific calculation. Observers are weighing the significance of the claims and what they mean for both the company and the field of algorithm design.