Martin Cid Magazine

DeepSeek V4.1 Flash: 552B parameters, 8B active, 60% cached-input cut — ties Opus 5

DeepSeek's newest model activates just 8 billion of its 552 billion mixture-of-experts parameters per token — a new Causal ...
Semiconductor Engineering

Heterogeneous System With Specialized HW For Disaggregated LLM Inference (Princeton Univ., Univ. of Washington)

A new technical paper titled “SPAD: Specialized Prefill and Decode Hardware for Disaggregated LLM Inference” was published by researchers at Princeton University and University of Washington. “Large ...
Semiconductor Engineering

Packet-Based NPUs In The LLM Era: From Compute-Bound CNNs To Memory-Bound Edge And Automotive Workloads

Many Semiconductor Engineering readers know the basic story behind Expedera’s Origin NPU IP architecture: packets instead of layers, higher MAC utilization, and less gratuitous movement of activations ...
AppleInsider

Apple AI research shows how MLLMs understand, generate, search for images

Apple's researchers continue to focus on multimodal LLMs, with studies exploring their use for image generation, understanding, and multi-turn web searches with cropped images. Now, the company is ...