Running Models Locally: Ollama, LM Studio and llama.cpp
For developers, analysts and IT teams who want to run open-weight models on laptops, workstations and small servers, and to do it securely enough for real work.
The curriculum is being prepared and reviewed. Enrollment is not open yet.
About this course
Local inference has become practical for everyday work: a recent laptop or a single workstation GPU can run capable open-weight models at interactive speeds, with no per-token bill and no data leaving the device. The three tools most people use sit on a shared foundation. llama.cpp (whose ggml team joined Hugging Face in February 2026) provides the GGUF format and the inference engine; Ollama wraps a model library, a Modelfile format and an API around it; LM Studio adds a desktop app, an MLX engine on Apple silicon and a headless daemon called llmster.
This course teaches you to choose models and quantizations that fit your hardware, build and tune llama.cpp and llama-server, manage models and Modelfiles in Ollama, and run LM Studio both on the desktop and headless. You will connect local models to coding assistants, RAG and MCP clients through OpenAI-compatible APIs, get reliable tool calls and structured output from smaller models, and run embeddings locally. The final module covers what can still leave your machine (cloud model options, telemetry, remote access), securing a model server that others can reach, license obligations, and when to move from a desktop tool to a shared server such as vLLM. Tool versions and features are as of October 2026.
What you will be able to do
Estimate whether a model and quantization will fit in available VRAM or unified memory and predict usable context length
Build and run llama.cpp and llama-server with appropriate GPU offload, context and parallel-slot settings
Choose a GGUF quantization type for a quality and memory budget and explain the role of an importance matrix
Manage models in Ollama, including custom Modelfiles and imported GGUF files, and call its API
Run LM Studio interactively and headless with llmster and the lms CLI
Connect local models to coding tools, RAG pipelines and MCP clients through OpenAI-compatible endpoints
Secure a local model server and identify which features send data off the machine
Syllabus
6 modules · 18 lessons · about 10 hours
1
Module 1: Planning a Local Setup
Why run locally, what hardware you need, and how to read a model listing.3 lessons
Why Run Locally: Privacy, Cost, Latency and Offline Use
Sizing Hardware: VRAM, Unified Memory, KV Cache and Tokens per Second
Reading a Model Listing: Parameters, Active Parameters, Quantization and Context Length
2
Module 2: llama.cpp: The Engine Underneath
Build it, choose quantizations and serve with llama-server.3 lessons
Building llama.cpp for CUDA, Metal, Vulkan and CPU
GGUF Quantization Types: K-Quants, I-Quants and Importance Matrices
llama-server: OpenAI-Compatible Serving, Parallel Slots and Context Settings
3
Module 3: Ollama
The model library, Modelfiles and the API.3 lessons
Pulling, Running and Managing Models with the Ollama CLI and App
Modelfiles: System Prompts, Parameters and Importing GGUF
The Ollama API, OpenAI Compatibility and Cloud Models: What Leaves Your Machine
4
Module 4: LM Studio
Desktop discovery, Apple silicon performance and headless operation.3 lessons
LM Studio Desktop: Model Discovery and the llama.cpp and MLX Engines
Headless LM Studio with llmster and the lms CLI
LM Studio's Local Server and Developer SDKs
5
Module 5: Putting Local Models to Work
Wire local models into the tools people already use.3 lessons
Connecting Local Models to Coding Assistants and MCP Clients
Reliable Tool Calling and Structured Output from Small Models
Local Embeddings and Private Document Search
6
Module 6: From Personal to Team Use
Security, licensing and knowing when to graduate.3 lessons
Securing a Model Server: Bind Addresses, Authentication and Network Exposure
Model Licenses, Provenance and Acceptable-Use Terms
When to Move to vLLM or a Shared Inference Server
Exam and certificate
The exam unlocks once you have completed every lesson. It draws 20 questions from a larger bank, spread across the modules, with the answer options shuffled for each attempt. You have 40 minutes and need 80% to pass. After a failed attempt you can try again after 24 hours, up to 3 attempts in total. Feedback is given per module, never as the answer key.
Passing earns a certificate of completion and assessment from the Wirewalk Institute of AI with a unique ID that anyone can check at institute.wirewalk.com/verify. It is not an accredited qualification or a vendor certification. How certification works.
The Wirewalk Institute of AI is independent. This course teaches publicly documented products and is not affiliated with or endorsed by Ollama, LM Studio, ggml (llama.cpp), Hugging Face, Apple (MLX), or NVIDIA (CUDA). Product and company names are trademarks of their respective owners.
A hands-on hour with ChatGPT, Claude and Gemini, told through one small business racing a deadline. Set up an account with a privacy setting you chose, decide what never goes into a personal account, write prompts that produce usable drafts, use files and web search, catch confident mistakes before they go out, and pick the assistant that fits how you work.
A practical, current tour of ChatGPT as a professional tool: choosing a plan and model, prompting well, analyzing files, researching with citations, using images and voice, organizing work in projects and memory, connecting apps and plugins, delegating multi-step work to ChatGPT Work, and knowing which admin and privacy controls matter. Ends with clear guidance on moving a workflow to the OpenAI API.
Learn Claude by solving one firm's real problem: getting through renewal season with too few analysts and without putting client data in the wrong place. Each module compares the approaches professionals actually choose between, shows a realistic failure and its fix, and ends with a Decision Lab where you make the call under pressure.