Getting Started: Vibe Coding Locally with AMD Ryzen™ AI and Radeon™ Graphics Cards

The way software is written is undergoing a fundamental change: as large language models grow ever more sophisticated – anyone can program and everyone is a coder. AI pair coding, more informally referred to as “Vibe Coding”, represents this next stage – where intent, creativity and system intelligence work in concert to translate ideas directly into functioning programs.

While competent coding models were previously the domain of cloud-based endpoints, over the past few months, models specialized in tool-calling, MCP and coding ability have appeared. Thanks to tools like LM Studio, Cline and Microsoft VS Code, you can get started with vibe coding in just a few minutes (depending on your internet speed). From there – you can just prompt your computer – and let the AI code for you. 

Running vibe coding locally on AMD hardware brings unique advantages. By processing everything on-device, sensitive code and data stay on your machine. Local execution also saves on recurring cloud costs, letting you scale experimentation without worrying about usage bills. Together, these benefits make AMD Ryzen™ AI and Radeon™ platforms the ideal foundation for practical, everyday vibe coding.

The AMD Ryzen™ AI Max+ 395 (128GB) is the best windows AI PC processor for local vibe coding in thin and light - unlocking the capability to run state-of-the-art models like GLM 4.5 Air (106 Billion Total Parameters) on Windows.

Cline is an open source AI coding bot that integrates into Microsoft Visual Studio Code, can access various MCP tools, and claims SOC2 and GDPR compliance. It is one of the most popular tools for vibe coders and senior developers alike. Powered by an LLM provider (in this case LM Studio), Cline allows a large language model to utilize the Microsoft VS Code GUI and even read/write files and folders on your computer. When equipped with a competent coding model like Qwen3-Coder 30B A3B or GLM 4.5 Air, it excels at autonomous coding tasks and AI pair coding. 

How to Vibe Code on AMD with Cline, LM Studio and Microsoft VS Code

1. Install the latest AMD Software: Adrenalin Edition 25.9.1 or higher drivers to get the latest llama.cpp support. For commercial customers, such as the HP ZBook Ultra G1a or HP Z2 Mini G1a please download AMD Software: PRO Edition 25.Q3 or higher. Note: models may fail to load or may not be performant for older drivers. 
2. AMD Radeon™ and Radeon™ PRO graphics card users can ignore this section and skip to #3. If you are using an AMD Ryzen™ AI platform, please set AMD Variable Graphics Memory (VGM) to the correct setting by consulting the product support matrix below. You can access the VGM control panel by right clicking on desktop > AMD Software: Adrenalin Edition > Performance > Tuning. If you are on an AMD Ryzen™ AI PRO platform such as the HP ZBook Ultra G1a or HP Z2 Mini G1a, please reboot to BIOS and set Dedicated Graphics Memory from there. Restart.
3. Download and install LM Studio. Skip onboarding. 
4. Go to the search tab (magnifying glass) and search for the correct model by consulting the product support matrix given below. We recommend using “LM Studio Community models” as the model vendor. You can select the quantization of the model on the right-hand side by clicking "Download Options" drop down menu on the right hand side (after selecting the model on the left hand side).
5. Click on the server tab (command line icon). 
6. Click on the drop-down menu at the top. Check “manually select parameters”. Click Load. Check "show advanced settings".
7. Set context length by referring to the product support matrix tables below. Check Flash Attention. This step is very important because without this – high context will not load in llama.cpp. Check “remember settings”. 
8. Click load. Minimize the LM Studio window (do not close it). 
9. Download and install Git for Windows. 
10. Download and install VS Code for Windows. 
11. Go the Cline Website and click on Install Cine for VS Code. 
12. In the LLM providers, select “LM Studio”. 
13. In VS Code, click on the cline icon on the right hand side and click on settings (wheel nut) and under Plan Mode and Act Mode (API configuration) make sure it is set to the model you downloaded in LM Studio such as Qwen 3 Coder. You may want to turn on Compact Prompts by consulting the product support matrix below.
14. Without exiting the Cline settings context menu, now click on “terminal settings” (on the left side) – which is a command line icon. Change the “Default Terminal Profile” dropdown to “Git Bash”
15. Click Done to close the Cline settings and save them. Start a new “Plan” task using the sample prompt given below and once you have a response from the Cline.bot. Shift the toggle to “Act” Mode for it to start working!

We are excited to continue innovating with our ecosystem partners and bring ever more sophisticated and performant models to AMD customers in the future – so the model of choice (for vibe coding) will likely change in the future.  We also expect product compatibility to grow in the future as smarter models lower the memory requirements for minimum viable competency of coding models. 

Resrouces