Articles by tag: llm-inference

Apple M5 Chip: 3nm Process, 128-Core Neural Block and AI Revolution

Apple M5: TSMC 3nm Process, 128 Neural Engine Cores, Zero-Copy Memory, and LLM Launch on a Laptop. Learn why M5 is the most aggressive attack on NVIDIA and Intel. Read details.

Qwen 3.5 397B on iPhone 17 Pro: Flash-MoE

Running 397 billion parameters Qwen3.5 on iPhone 17 Pro with SSD and Flash-MoE. Speed 0.6 t/s, comparison with M3 Max. Technical details for developers.

vLLM Production Stack: basic launch in K8s

Deployment of vLLM in Kubernetes with Production Stack. GPU/CPU settings, tool-calling, multimodal models. YAML configurations and memory optimizations for middle/senior devops.