Aleph Alpha's Kolibri-1, a 78-billion-parameter mixture-of-experts model with about 3.46B active parameters per token, requires roughly 78GB of FP8 weights and real-world serving of about 140GB of VRA…
NVIDIA Corporation released an interactive 3D visualization of its H100 GPU architecture using Three.js, allowing users to explore components such as CUDA cores, HBM3 memory, GPCs, SMs, Tensor Cores, …
A production configuration for running DeepSeek-V4-Flash-0731 on a single AMD MI300X GPU achieves 168.6 tok/s median single-stream decode and 542 tok/s aggregate across 8 concurrent streams, with the …
Meta researchers developed In-Kernel Broadcast Optimization (IKBO), a kernel-model-system co-design that eliminates redundant user-embedding replication during recommendation model inference. Deployed…