On the 28th of January 2026, a small team in New York shipped a new version of an app that lets you run powerful AI models on your own laptop, without sending a single piece of data to the cloud. The app is called LM Studio. It is free for personal use, free for use at work, and the latest release turns it into something that organisations can deploy properly. So today I am asking, is it time you tried running AI on your own machine?
Before I get into LM Studio specifically, let me explain what running a local AI model actually means. When you use ChatGPT, Claude, or Gemini, you are sending your prompts and your data to someone else’s servers. But if you are working with sensitive client information, draft contracts, internal HR queries, or anything you would not want to leave your organisation, you have a problem. Running a model locally means the AI lives on your laptop. The model, the prompt, and the response all stay on your hardware. Nothing goes out. That is the pitch.
What LM Studio Actually Does
LM Studio is a desktop app made by a small team at Element Labs in New York, founded by Yagil Burowski, who used to work at Apple. It launched in May 2023, and on the 8th of July 2025, the team announced that LM Studio is free for use at work. No commercial licence. No form to fill in. Your team can just install it and use it.
The app gives you a clean interface where you can browse and download open models from Hugging Face, the world’s largest model repository. So we are talking about Meta’s Llama, Mistral, Google’s Gemma, Alibaba’s Qwen, Nvidia’s Nemotron, OpenAI’s Open GPT-oss, DeepSeek, and many more. You pick one, you download it, and you chat with it the same way you chat with ChatGPT, except it is running on your machine.
It also has a built-in retrieval feature, which means you can drop your own documents in and ask questions about them. Your contracts, your reports, your style guides, none of which leave your laptop. For anyone in regulated work, that is a fundamentally different proposition from uploading a confidential file to a chatbot.
The Upgrade
The release of LM Studio version 0.4.0 is what makes this an episode worth doing now. Three changes matter.
The first is something called LLMster. It is the LM Studio engine, packaged to run without the desktop interface, on a server. That sounds technical, but the implication is simple. An IT team can now run LM Studio for a whole department, on shared infrastructure, the way they would run any other internal tool.
The second is parallel requests. Previously, if 5 people queried the same local model, they queued up. Now they can all be served at the same time. That is the difference between a personal toy and something a team can actually use.
The third is a new API endpoint that mirrors how OpenAI’s API works. So if your developers have already built workflows that talk to ChatGPT, they can point those workflows at a local model with very little change. Same plumbing, different brain.
There is also an Enterprise plan with single sign-on and the ability to control which models people in your organisation are allowed to run. That is the kind of feature large organisations need before they will let anything near a corporate laptop.
Governance and what it takes to power it
This is not a free lunch. Local AI runs on your hardware, which means your hardware has to be capable. LM Studio recommends at least 16 gigabytes of RAM, and ideally a graphics card with 4 gigabytes of dedicated memory or more. Apple Silicon Macs from M1 onwards are particularly well suited because of how their memory is designed. Windows machines need a modern processor and a decent GPU. Linux is supported, including on the new Nvidia DGX Spark and DGX Station boxes.
If you are working on a 5-year-old laptop with 8 gigabytes of RAM, this is not for you yet. The smaller models will run, but slowly. The bigger, more capable models need serious kit. So the trade-off is this. You stop paying a monthly subscription to a cloud provider, but you pay once for hardware that can keep up. For an individual, maths is borderline. For an organisation handling sensitive data, it makes a lot more sense.
Back in episode 308, I covered France’s deal with Mistral, where the French military insisted that every AI system runs on French infrastructure, under French control. That is sovereign AI at a national level. LM Studio is sovereign AI at a personal and organisational level. Same idea, different scale.
And it connects to something I have been talking about a lot, the rise of domain expertise AI. I am building one myself, GovComms, for government and public sector communications teams. The point of these specialist tools is that they understand your context. The point of local AI is that they live where your data lives. Put the two together, and you start to see the shape of what serious AI use will look like in 2026 and beyond. Specialist, private, on infrastructure you control.
If your organisation is still copying confidential text into a public chatbot, you are doing it wrong. LM Studio is not the only answer, but it is one of the easiest ways to start doing it right.
If you want to build your AI literacy and skills, I can help
📚 Grab AI Playbook on Amazon, the perfect podcast companion.
📰 Subscribe to Human in the Loop, my LinkedIn Newsletter
🎓 Take the CPD Accredited AI Playbook Diploma
📞 Book a call at digitaltraining.ie
Sources
- LM Studio is free for use at work (LM Studio blog, 8 July 2025) — https://lmstudio.ai/blog/free-for-work
- Introducing LM Studio 0.4.0 (LM Studio blog, 28 January 2026) — https://lmstudio.ai/blog/0.4.0
- LM Studio Blog (full release timeline) — https://lmstudio.ai/blog
- LM Studio System Requirements (official docs) — https://lmstudio.ai/docs/app/system-requirements
- LM Studio Enterprise plan — https://lmstudio.ai/enterprise
- Simon Willison on LM Studio free for work (8 July 2025) — https://simonwillison.net/2025/Jul/8/lm-studio-is-free-for-use-at-work/
- LM Studio launches NVIDIA DGX Spark support (October 2025) — https://lmstudio.ai/blog/dgx-spark
- LM Studio supports NVIDIA DGX Station GB300 (March 2026) — https://lmstudio.ai/blog/dgx-station
- Element Labs $19.3M funding (AlleyWatch, June 2025) —https://www.alleywatch.com/2025/06/the-alleywatch-startup-daily-funding-report-6-2-2025/
- AI SIX Episode 308 — France Chooses Sovereign AI —https://aisix.buzzsprout.com/
Key Statistics
- Founded: May 2023 by Yagil Burowski, a former Apple engineer, at Element Labs in New York
- Free for personal use since launch
- Free for commercial / work use as of 8 July 2025
- Latest major version: LM Studio 0.4.0, released 28 January 2026
- New in 0.4.0: llmster headless daemon, parallel requests with continuous batching, new stateful /v1/chat REST API, refreshed UI
- Recommended hardware: 16GB+ RAM, GPU with 4GB+ VRAM (Apple Silicon M1 or later, Windows AVX2-capable processors, Linux x64 and ARM)
- Built on llama.cpp and Apple MLX inference engines
- Built-in Hugging Face model browser; supports Llama, Mistral, Gemma, Qwen, Nemotron, gpt-oss, DeepSeek and many others
- Built-in RAG (Retrieval-Augmented Generation) for chatting with your own documents
- Enterprise plan available with SSO and model gating
- Cross-references: AI SIX Ep 308 (France Chooses Sovereign AI), Ep 280 (A Domain Expertise LLM)
About the Podcast Host
AI and Communications Specialist | Author | Podcaster | International Speaker
Joanne Sweeney is a recognised leader in digital marketing and digital transformation, specialising in the practical application of AI to enhance communication strategies. The founder of Digital Training Institute, Joanne, empowers leaders and communications professionals to leverage AI responsibly and effectively, ensuring transparency, trust, and tangible results.
She completed the Harvard Kennedy School Executive Programme in AI and Public Policy this summer before launching her fourth book AI Playbook, in September. Joanne has also published two editions of Public Sector Marketing Pro, which were preceded by her first title, Social Media Under Investigation Law Enforcement and the Social Web.
Her books have become go-to resources for governments, public sector and policing agencies, as well as marketing and communications professionals, all navigating the challenges of the changing Internet.
Joanne is the host of the AI SIX Podcast, a daily six-minute podcast that provides actionable insights, expert interviews, and practical advice to help professionals navigate the new world of work.
