THE CONNECTED LIBRARY

Explore your next question.

Start anywhere. An explanation can lead to an experiment, a tool, or a different way to solve your problem.

A useful next step, wherever you start.

Explanations, practical guides, tools and news share the same territory. Choose a topic to see how they connect.

23 connected resources
Tool

Local LLM memory calculator

Change the model, hardware and context. See where the memory goes.

Try it
Reference

Model explorer

Inspect model sizes, capabilities, source records and memory estimates.

Explore
Reference

Hardware explorer

Compare GPUs and unified-memory computers for a local workload.

Explore
Article

How to write a prompt that gets useful work done

Turn a vague request into a clear task, give the AI the right material, and judge the result against something concrete.

Explore
Article

How to give an AI the context it actually needs

Choose what to include in a prompt, make sources easy to distinguish, and diagnose answers that overlook your documents.

Explore
Article

How to check an AI answer before you use it

Separate factual claims, calculations and suggestions, then verify each with evidence suited to the task.

Explore
Guide

How much VRAM do you actually need?

Calculate model weights, KV cache and runtime overhead before choosing hardware for local AI.

Explore
Guide

What is quantization?

Understand Q4, Q5, Q8 and FP16, and how lower precision trades model size for possible quality changes.

Explore
Guide

How context uses memory

See how context length, the KV cache and concurrent requests affect local LLM memory requirements.

Explore
Guide

Mac or GPU workstation for local AI?

Compare unified memory and discrete GPU memory, runtime compatibility and upgrade paths for local models.

Explore
Guide

Two GPUs do not become one bigger GPU.

Understand the conditions under which multiple GPUs can run a larger local model, and what combined VRAM leaves out.

Explore
Guide

Buying a used RTX 3090

A practical checklist for assessing a used 24 GB GPU for local AI, including memory, thermals, power and seller evidence.

Explore
Guide

Why is my local LLM so slow?

Troubleshoot slow Ollama, LM Studio and llama.cpp inference. Separate loading, prompt processing and generation before changing hardware.

Explore
Guide

Why your local LLM runs out of memory in longer chats

Fix local LLM out-of-memory errors as context grows. Budget the KV cache, reduce concurrency and test cache quantization without losing key information.

Explore
Guide

Smaller Q8 or larger Q4: which local model should you run?

Choose between a smaller high-precision LLM and a larger quantized model with a practical test for quality, memory, context and response time.

Explore
Guide

Why the same model gives different answers across local AI apps

Compare Ollama, LM Studio and llama.cpp fairly. Check the model artifact, chat template, sampling, thinking mode and context before blaming the model.

Explore
Guide

Local RAG giving bad answers? Check retrieval first.

Fix local document Q&A with a practical RAG checklist: PDF extraction, chunks, embeddings, hybrid search, reranking and a small evaluation set.

Explore
Guide

Ollama hardware requirements: how much RAM and VRAM do you need?

Plan Ollama RAM and VRAM for 8GB, 12GB, 16GB and 24GB GPUs. Compare worked model-memory examples and check GPU offloading before buying hardware.

Explore
Guide

Ollama vs LM Studio: which should you use for local AI?

Compare Ollama and LM Studio for local chat, APIs, offline use and Mac runtimes. Choose by workflow, then test the same model and memory settings.

Explore
Guide

How much Mac memory do you need for local LLMs?

Compare 16GB, 24GB, 32GB and 64GB Mac memory for local LLMs. Budget unified memory, context and macOS headroom with worked model examples.

Explore
News

Qwen3.8: long context starts with a memory plan

A practical look at the Qwen3.8 27B catalog entry: stored weights, context settings and what to check before a local run.

Explore
News

Gemma 4 E2B: read beyond the small-model label

The stored checkpoint matters when planning memory. Here is how to turn a compact-model headline into a useful hardware check.

Explore
News

Mixture of experts: active parameters are only part of the story

Gemma 4 26B-A4B is a useful example of why compute-oriented model labels and a local memory budget answer different questions.

Explore