LM Studio has been my go-to app for working native LLMs since I found it. It is simple to make use of, the UI is well-polished, and you’ll have any AI mannequin up and working in a matter of some clicks — even for those who’ve by no means touched a terminal in your life. It is top-of-the-line instruments if you wish to get pleasure from the advantages of a neighborhood LLM.
However whereas LM Studio is free to make use of, it is not open-source. A few of its elements, particularly the command-line tooling, carry open licenses, however the app itself is proprietary software program. For a software that may simply sit on the core of your native AI workflows, risking a license change every time the mum or dad corporations need is a danger I used to be uncomfortable with, after which I discovered Jan.
Associated
I’ll by no means pay for AI once more
AI doesn’t must value you a dime—native fashions are quick, non-public, and eventually price switching to.
I believed I’d miss LM Studio. I didn’t
Why Jan looks like an actual substitute for LM Studio
The choice that finally ended up changing LM Studio for me is Jan. It is a desktop software that allows you to run LLMs totally offline — very like LM Studio, however not solely is it utterly free, it is also open-source, with all of the supply code being accessible on GitHub. There are not any licensing surprises, no proprietary lock-ins, and no effective print you could fear about.
The interface is so much like ChatGPT, which can be a very good or dangerous factor relying in your style. If you happen to’re new to the idea of native LLMs, although, it is acquainted, clear, and welcoming. Jan was made to make native AI accessible to everybody, so the UI design imitating ChatGPT is not an accident. You get a chat window, a mannequin hub, and a settings web page that does not require developer-level experience to grasp.
If you happen to’re switching from LM Studio, the mannequin library will really feel acquainted too. You’ll be able to browse and obtain well-liked open-source fashions like Llama, Gemma, Mistral, Qwen, DeepSeek, and extra instantly from inside the app. Fashions are additionally tagged to inform you whether or not they’ll run effectively in your {hardware} or if they will be a bit an excessive amount of to deal with.
Yadullah Abidi / MakeUseOf
Jan additionally comes with an OpenAI-compatible API server. This implies you can level different instruments like Cursor, Open WebUI, customized scripts, and no matter else you may be constructing at Jan’s native server and have it run simply as it will when speaking with OpenAI’s API. That is nice to check out code and prototypes earlier than you begin interfacing with OpenAI’s precise (and paid) API. The server additionally has CORS (Cross-Origin Useful resource Sharing) assist by default, which makes it fairly simple to hook as much as net tasks.
Efficiency, though dictated by the mannequin you are utilizing, can be fairly much like LM Studio. Token utilization is not considerably completely different from LM Studio both, and the uncooked inference engine is basically the identical. You additionally get built-in extensions in case you wish to prolong the app’s performance.
OS
Home windows, macOS, Linux
Developer
Jan
Worth mannequin
Free, Open-source
A free, open-source AI chat assistant that runs native giant language fashions in your PC, no cloud required.
The privateness advantages are literally significant
Native fashions, no cloud, no surprises
Yadullah Abidi / MakeUseOf
Working AI domestically is already the privacy-conscious alternative; nonetheless, utilizing Jan enables you to go a step additional. All the pieces out of your preferences, chat historical past, mannequin parameters, and the mannequin itself stays in your machine. There is no account to create, no telemetry information to fret about, and no cloud dependency required. You’ll be able to have a completely airgapped AI consumer expertise in order for you it.
Two drawbacks of this method are that first, you are restricted to the AI fashions your {hardware} can run, and second, open-source LLMs you can obtain and run aren’t at all times pretty much as good as their on-line counterparts. There are duties that it is best to use a neighborhood LLM for, however it could actually’t do every little thing.
Jan enables you to treatment this by supporting distant APIs like OpenAI, Anthropic, Claude, and extra in order for you cloud entry to their respective fashions. This method does not require you to decide on between the 2. It is native (and personal) by default, however you at all times have the choice to decide on cloud-based fashions when the duty at hand requires one.
It’s not flawless
The tough edges you’ll discover rapidly
Yadullah Abidi / MakeUseOf
If you happen to’re switching over from LM Studio, you are going to miss the element it brings, particularly in GPU layer configuration, which is way extra visible and accessible in LM Studio. Moreover, options like doc evaluation are extra mature. Jan has been catching up, however in order for you fine-tuning management for each inference parameter, LM Studio is the higher alternative.
If you happen to choose working within the terminal, Ollama beats each LM Studio and Jan. As such, startup instances for each visible LLM applications are comparable, however slower than Ollama. Jan sits someplace in the course of the competitors — extra clear than LM Studio and considerably extra user-friendly than Ollama.
It quietly turned my default
Why I finished opening LM Studio altogether
There’s not a lot stopping you from shifting to Jan from LM Studio. The educational curve is sort of flat, and most of your fashions and information will carry over cleanly. The distinction is in possession of your native AI stack. With Jan, you are in full management, and no surprising pivots or licensing adjustments will catch you off guard.
Associated
I now use this offline AI assistant as an alternative of cloud chatbots
Even with cloud-based chatbots, I will at all times use this offline AI assistant I discovered.
In sensible, each day use, Jan does every little thing LM studio does, besides it comes with the peace of thoughts that comes from software program you may truly audit. You continue to get all of the management and benefits you want, simply with none potential problem.
