Learn · Local AI path

From concept
to a measured runtime.

Understand the model and hardware constraints, run a model with Ollama, then inspect placement, memory and measured behavior instead of stopping at installation.

Step 01

What is Local AI?

Understand what runs locally and what changes when inference moves to your own hardware.

Step 02

Hardware for Local AI

Learn which CPU, RAM, GPU and storage constraints matter.

Step 03

LLM Model Size

Understand parameters, precision and why parameter count is not a memory figure.

Step 04

VRAM Explained

Understand why model file size is not the same thing as runtime GPU memory.

Step 05

Quantization

See how lower-precision model weights change memory requirements and trade-offs.

Step 06

GGUF

Learn the model format commonly used by llama.cpp-based local workflows.

Step 07

Context Length

Separate model context limits from active context and understand the memory trade-off.

Step 08

CPU vs GPU

Understand CPU inference, full GPU residency and partial GPU offload.

Step 09

Ollama

Understand the runtime, model management and local API.

Step 10

Install Ollama on Windows

Follow the tested tutorial from installation through a local API request.

Step 11

Run a local model

Move from installation to a repeatable model-running workflow.

Step 12

Check model fit

Estimate whether a model is likely to fit GPU VRAM, require partial offload or rely heavily on system RAM.

Step 13

Inspect Ollama memory

Read runtime size, size_vram, context and processor placement without conflating the fields.

Step 14

Understand GPU offloading

Interpret full-GPU and mixed CPU/GPU placement and the limits of placement percentages.

Step 15

Validate with measured results

Compare the planning model with reproducible Local AI measurements and explicit limitations.

USE THE PATH AS A LOOP.
If the runtime does not fit or perform as expected, go back to model size, quantization, context or hardware constraints and test another configuration.