All articles
Guides, experiments and analysis for running AI on your own hardware.
- Deepseek v4.1 Flash 1017 Visual Review: A Natural Artist · Model analysis
- Quantization in Pictures: Fewer Bits, Fewer Colours · Fundamentals
- Why Local LLMs Slow Down as Conversations Grow · Fundamentals
- DeepSeek V4.1 Flash: 552B Architecture, Engram Memory, and the 890 B/Token KV Cache · Model analysis
- GLM 5.3 Flash: My Experience Using It · Model analysis
- Apple M5 Ultra for Local AI · Hardware
- Qwen 3.8 27B: The Practical Sweet Spot for Local AI · Model analysis
- Qwen 3.8 Flash Explained: The 51B N-Gram Memory Architecture · Model analysis
- Long-Context and Attention in LLMs Explained (RoPE, YaRN) · Fundamentals
- I Built a 48 GB VRAM Local AI Workstation with Two RTX 3090s · Hardware
- How to Access Your Local AI From Anywhere · Software
- GPU Offloading for Local LLMs · Fundamentals
- Ollama, llama.cpp or vLLM: Which One Should You Use? · Software
- Reasoning Models and Thinking Tokens Explained · Fundamentals
- Speculative Decoding and MTP Explained · Fundamentals
- Kimi K3 Is a 2.8-Trillion-Parameter Open-Weight Monster · Model analysis
- LLM Quantization Without the Alphabet Soup · Fundamentals
- The Best Affordable Multi-GPU Motherboard for Local AI · Hardware
- DeepSeek V4 Flash Is Almost Too Cheap to Meter · Model analysis
- Apple Silicon Macs for Local LLMs · Hardware
- NVIDIA DGX Spark and GB10 for Local LLMs · Hardware
- AMD Strix Halo for Local LLMs · Hardware
- Best GPUs and Devices for Local AI Inference · Buying guide
- RTX 3090 for Local AI Inference · Benchmark