marksmith.dev
My Thoughts
Exploring web development, design patterns, and the future of technology.
Tag: #LLM
Why Your Old Xeon Is Still Painfully Slow for LLMs — AVX Explained
AVX, AVX2, and AVX-512 are not marketing buzzwords. They are the difference between an LLM that spits out text in real time and one that sits there grinding through a single prompt for minutes. Here is why your older hardware is holding you back, even if you bolted a modern GPU onto it.
Running LLMs Locally vs Renting Cloud GPUs: The Real Cost Breakdown
As of April 8, 2026, the current official numbers are stark. AWS On-Demand shared GPU capacity ranges from $10.49264/hour for g6e.12xlarge to $55.04/hour for p5.48xlarge, while current public marketplace listings from RunPod and Lambda sit much lower for narrower GPU slices. The right answer is still utilisation: rent for bursts, buy for steady workloads that fit your hardware.
Artificial Analysis: The Independent AI Benchmarking Platform You Need to Know
With AI models dropping faster than most teams can evaluate them, Artificial Analysis has carved out a rare niche: independent, rigorous, and genuinely useful benchmarking that cuts through the hype.
Qwen 3.6 Plus: Alibaba's Flagship AI Model Redraws the Agentic Coding Boundary
Qwen 3.6 Plus launched April 2 2026 with a 1 million token context window, a 61.6 score on Terminal-Bench 2.0, and a pricing model that undercuts Claude Code at roughly $0.50 per million tokens on Bailian. The agentic coding race just got significantly more competitive.