LM Studio

Good for / not for

What it actually does

LM Studio is the graphical front door to local models. You browse a catalogue, download a model and chat with it in a window — no terminal, no config file. The free tier covers local inference and voice transcription on your machine, the Bionic Agent, web search, and LM Link across up to five devices. It can use Vulkan GPU offload on AMD and Intel laptops, which is the practical reason it shows up on machines without an Nvidia card. Two things set it apart from the rest of this category. First, it is not open source: LM Studio's terms state that the source code constitutes trade secrets, so there is nothing to fork or audit, and the company does not disclose download or user counts, which leaves no independent measure of its reach. Second, since 2026 there has been a paid layer above the free one. Cloud credits are pay-as-you-go per million tokens — DeepSeek V4 Flash at $0.13 in, $0.028 cached and $0.26 out, Kimi K3 at $3.00, $0.30 and $15.00 — and a subscription called Bionic Pass has been announced but is not yet purchasable, so there is no price to quote for it. Local use stays free.

Pricing and free-plan limits

Free tier: Local inference is unlimited and LM Link covers up to five devices. Cloud models need paid credits, with no free allotment stated.

Three honest limitations

1. A review puts the requirement at 16GB of RAM minimum and 32GB recommended, with model downloads ranging from 2GB to over 100GB.
2. The same review notes you are limited to open-weight Hugging Face models unless you pay for cloud credits.
3. It is proprietary, with no repository to inspect, and download and user counts are not disclosed — so there is no independent measure of how widely it is used.

Alternatives

• Jan — Choose Jan instead if you want the same desktop shape under an open licence, with no paid tier and no cloud credits to think about.
• Ollama — Choose Ollama instead if a terminal is comfortable and you want an OpenAI-compatible local API to build against.
• AnythingLLM — Choose AnythingLLM instead if the job is chatting with your own documents rather than picking and running models.

Getting started in 5 minutes

1. Check the machine first: a review puts the floor at 16GB of RAM, with 32GB recommended.
2. Download LM Studio from lmstudio.ai. The terms make it free for personal and internal business use.
3. Pick one open-weight model from the built-in catalogue and chat with it before downloading a second — the files run from 2GB to over 100GB.
4. On an AMD or Intel laptop, use the Vulkan GPU offload the app supports rather than assuming you need an Nvidia card.
5. Buy cloud credits only if you need models that will not run locally. They are priced per million tokens, and the announced Bionic Pass subscription cannot be bought yet.

LM Studio questions

Is LM Studio free?

For local use, yes — and the terms grant that for personal and internal business use, not personal use alone. Local inference and voice transcription, the Bionic Agent, web search and LM Link across up to five devices are all included. The cloud side is what costs: those models are billed as pay-as-you-go credits per million tokens.

Is LM Studio open source?

No. It is proprietary, and its terms state that the source code constitutes trade secrets. That is the sharpest difference from the rest of this category: Ollama and AnythingLLM are MIT, Jan is Apache 2.0. LM Studio still runs models locally — you simply cannot read or fork the app doing it.

How much RAM do I need?

A review of LM Studio puts the minimum at 16GB, with 32GB recommended, and notes model downloads from 2GB to over 100GB. Plan for disk as well as memory once you keep more than one model around.

How many people use it?

There is no public number. LM Studio is closed source, so there is no star count, and download and user counts are not disclosed. What is on the record is coverage: Windows Central and InfoWorld have both written it up as the leading graphical alternative to Ollama.

What is the Bionic Pass?

An announced subscription that is not yet purchasable, so there is no price to give for it. Today the only paid option is pay-as-you-go cloud credits.

Sources

• LM Studio pricing (free local tier, LM Link to 5 devices, cloud credit rates, Bionic Pass announced but not purchasable)
• LM Studio terms (proprietary; source code described as trade secrets; free for personal and internal business use)
• LM Studio review (16GB RAM minimum, 32GB recommended, 2GB–100GB+ downloads, open-weight Hugging Face models)
• Windows Central on LM Studio as the graphical alternative to Ollama
• InfoWorld: first look at running LLMs locally with LM Studio