Uncensored AI ChatWindows · macOS · Linux · API
LM Studio
Runs open models on your own hardware, free, with nothing leaving the machine.
What it does
- Downloads and runs open-weight models entirely on your own hardware
- Model catalogue browsable and installable inside the app
- Local server exposing an OpenAI-compatible API on your machine
- Voice transcription processed on device
At a glance
- Price
- On request
- Free tier
- Yes
- Account
- No
- Runs locally
- Yes
- Image generation
- No
- Voice
- Yes
- Trains on your content
- No
- Last verified
- 8 Aug 2026
ClickTug verdict
The honest answer to the search that brings most people to this category. Anyone typing AI with no restrictions is usually objecting to a hosted service refusing them, and running the model yourself removes the party doing the refusing. LM Studio is the most approachable way to do that: a real installer, a model browser, and a local API that existing tools already understand. The cost is hardware and the willingness to pick a model. In exchange there is no retention policy to read, because there is no retention.
Should you use LM Studio?
Good fit if
- Genuinely free and genuinely unlimited for local use, with no account
- Nothing leaves the machine, so there is no retention policy to trust
- Local OpenAI-compatible API, so existing tools work against your own hardware
- Works fully offline once a model is downloaded
- You choose the model, so refusal behaviour is yours to select
It is the strongest privacy answer available in this category, and it is not close. No account, no retention policy to trust, no terms that can change under you, and no filter applied by anyone. It is also the cheapest, since running a model you already downloaded costs electricity.
Look elsewhere if
- Needs decent hardware, and a weak machine makes a good model feel bad
- Text only. No image or video generation
- Model quality varies enormously and choosing well takes some learning
- The optional cloud tier is a hosted service under separate terms
- No support line, because there is no service
It demands hardware and a little patience. Model choice is on you, quality varies enormously between models, and there is no support line when a download behaves oddly. If you want something that works well the moment you open it, a hosted service will feel better.
What is LM Studio?
LM Studio is a desktop application that downloads open-weight language models and runs them on your own computer. It provides a chat interface, a model browser and a local server that speaks the OpenAI API, so existing tools can be pointed at your machine instead of at a provider. Once a model is downloaded it works with no internet connection at all.
Who it is for
Anyone whose objection to hosted assistants is that somebody else holds the conversation. It suits people who want to choose the model rather than accept whichever one a service offers, and developers who want an OpenAI-compatible endpoint without an account or a bill.
Who it is not for
Anyone without the hardware. Useful models want a reasonably modern machine with real memory, and a laptop that struggles will make a good model feel like a bad one. It also generates no images or video, so it is a text tool only.
Every fact we hold on LM Studio
| Category | Uncensored AI Chat |
|---|---|
| Pricing model | Freemium |
| Starting price | No recurring price published |
| Free tier | YesLocal use is free with no account and no usage cap. The paid tier is optional cloud inference, not a limit removal. |
| Free trial | No |
| Card required at signup | No |
| Account required | No |
| Payment methods | Card |
| Platforms | Windows, macOS, Linux, API |
| Generation modes | Text chat, Text generation, Text to speech |
| Content filtering | No documented output filter |
| Age policy | None stated |
| Position on minors | Not applicableRuns on your hardware and receives no user content. |
| Data retention | For local use there is nothing to retain: models run on your machine and LM Studio states no data ever leaves the device. The optional cloud inference is described as United States based with zero data retention.source |
| Trains on your content | No |
| Deletion on request | Your own files. There is no account and no vendor-held copy to delete |
| Anonymous use | Supported |
| Company | Element Labs, Inc., United States |
| Last verified | by ClickTugger |
Pricing
Local use is free with no cap and no account, and that is the actual product rather than a trial of it. LM Studio also sells optional hosted inference priced per token, from 0.13 to 3.00 US dollars per million input tokens and 0.26 to 15.00 per million output tokens depending on model. Paying is for convenience or for models too large to run at home, not to lift a restriction.
The free option is the whole thing. Local model use is unlimited, needs no account and has no trial period. The only paid element is optional cloud inference for models you cannot or would rather not run yourself.
| Tier | Price | Includes |
|---|---|---|
| Local | Free |
No account required. This is the product, not a trial of it. |
| Cloud, pay as you go | On request |
Input runs from $0.13 to $3.00 per million tokens and output from $0.26 to $15.00, depending on model. A separate hosted service under its own terms. |
Refunds: Not applicable to local use, which is free. Cloud credits are consumption-based.
Getting started
No, and no account is required at all for local use. You download the application and run it. That makes it one of very few tools in this directory that supports genuinely anonymous use.
- No card at signup
- No account required
- Anonymous use supported
Windows, macOS and Linux as a desktop application. It also exposes a local OpenAI-compatible API server, so anything that can talk to that endpoint can use your local models.
- Windows
- macOS
- Linux
- API
Privacy, data and safety
For local use the question mostly dissolves: models run on your hardware, LM Studio states that no data ever leaves your device, and voice transcription is processed on the machine. There is no server-side history because there is no server. The optional cloud inference is a separate hosted service, described as United States based with zero data retention, and using it means sending prompts to somebody else under different terms.
| Retention | For local use there is nothing to retain: models run on your machine and LM Studio states no data ever leaves the device. The optional cloud inference is described as United States based with zero data retention.source |
|---|---|
| Trains on your content | No |
| Deletion | Your own files. There is no account and no vendor-held copy to delete |
| Processing location | Your own device for local use |
| Privacy policy | Read the vendor policy |
Where your words go
Local mode
The primary mode, on your own hardware
Optional hosted mode
Only once you choose a remote provider
Cloud, pay as you go
A separate hosted service under its own terms
What that means for you
Yes. In its primary mode it runs models on your own hardware, so prompts and output stay on the machine and no vendor retention period applies to them. That holds until you use its optional hosted mode, which sends your prompts to a remote model provider under that provider's policy rather than this one. Cloud, pay as you go is a separate hosted service and none of that covers it.
Content policy
LM Studio documents its output filtering as no documented output filter, and its age policy as none stated.
LM Studio does not offer clothing removal from photographs of real people. ClickTug does not list tools that do. Its stated position on the likeness of real people is that it is blocked outright.
What the vendor prohibits
- Nothing is enforced by the application, because it runs locally and receives no content
Refusal behaviour is a property of whichever model you download, not of LM Studio. A permissive model will be permissive and a restricted one will refuse. The application does not filter either way.
LM Studio runs on your own hardware, so it receives none of what you generate. ClickTug does not require a minor safety policy or an acceptable use policy from a tool that never receives user content, because there would be nothing for such a policy to govern. Any optional hosted tier is a separate service under its own terms.
What the vendor publishes
This measures disclosure, not quality. A good product can document nothing, and that is worth knowing separately from whether it is any good.
| Price published | Yes, from On requestsource | Verified |
|---|---|---|
| Operating company named | Element Labs, Inc. | Documented |
| Jurisdiction stated | United States | Documented |
| Retention period stated | No period givensource | Not documented |
| Training policy stated | Yessource | Documented |
| Refund terms published | Yes | Documented |
| Content policy published | Self-hosted, not applicable | Not documented |
| Privacy policy published | Yessource | Documented |
| Position on minors published | Not applicable | Not documented |
Hands on
Most people who search for AI with no restrictions are not looking for adult content. They are looking for an assistant that will not stop mid-sentence to explain why it would rather not, on a subject that is legal, ordinary and frequently medical. The complaint is about refusals, and the durable answer to refusals is to remove the party doing the refusing.
That is what running a model locally does, and it is why this listing sits in a directory otherwise full of hosted services. LM Studio downloads an open-weight model onto your machine and runs it there. There is no account, no usage cap, no policy page governing what you may ask, and no server-side history, because there is no server. Whether the model refuses anything is a property of the model you chose, which means the decision is yours in a way it never is with a hosted product.
The trade is hardware. A model has to fit in memory before it can run at all, so the practical question is not how fast your machine is but how much memory it has. A modern laptop handles smaller models comfortably. Larger ones want a real GPU or an Apple Silicon machine with generous unified memory. A model that barely fits will run slowly enough to feel worse than it is, which is the commonest reason people try this once and conclude local models are poor.
Worth being precise about the paid tier, because it changes the privacy answer. LM Studio sells optional hosted inference billed per token, for models too large to run at home. Using it means sending prompts to somebody else’s computer under somebody else’s terms, and the fact that the application is otherwise local does not follow the data there. LM Studio describes that service as United States based with zero data retention. The local mode is the one this listing is about.
One note on why this is listed at all, since ClickTug normally requires a published minor safety policy before it will list anything. LM Studio does not publish one and does not need to: it receives no user content, so there is nothing such a policy could govern. That exemption is recorded explicitly against this listing rather than applied quietly, it covers only tools that run on your own hardware, and it does not extend to any hosted service, including this vendor’s own cloud tier.
How LM Studio compares
Every figure on the right is computed across the 41 tools ClickTug currently tracks, at the moment this page was built. They are facts about this catalogue rather than about the market, and they move as it grows.
| Measure | LM Studio | Across ClickTug |
|---|---|---|
| Starting price | Not published | $11.99 median of 14 that publish one |
| Publishes a price | No | 24 of 41 tracked |
| Free tier | Yes | 35 of 41 tracked |
| Names its operating company | Yes | 37 of 41 tracked |
| States a training position | Yes | 20 of 41 tracked |
LM Studio compared with the alternatives
Every cell is read from the same records as each tool's own listing, so nothing here can claim something its page denies. Rows needing an account to confirm are left out entirely rather than guessed at, and named in the verification log at the foot of the page.
| LM Studio | Venice AI | Ollama | |
|---|---|---|---|
| Starting price | Free | $18/mo | $20/mo |
| Free tier | Yes | Yes | Yes |
| Card to start | No | No | No |
| Account required | No | Yes | No |
| Runs on your hardware | Yes | No | Yes |
| Image generation | No | Yes | No |
| Video generation | No | Yes | No |
| Voice | Yes | Yes | No |
| API | Yes | Yes | Yes |
| Mobile app | No | No | No |
| Trains on your content | States no | Not stated | States no |
| Account deletion | Your own files | By request | Your own files |
| Company named | Yes | Yes | Yes |
| Position on minors | N/A | Published | N/A |
LM Studio vs Venice AIFreemium
Choose Venice instead if you want the reduced-refusal experience without owning the hardware, and can accept a hosted service holding the session.
LM Studio vs OllamaFreemium
Choose Ollama instead if you prefer a command line and a scriptable server over a desktop application, or want the same models from a terminal.
Quick answers about LM Studio
Generated from the same records as the tables above, so an answer here cannot contradict a figure there. Questions only appear when the answer says something specific to this product.
- Is LM Studio free?
- There is a free tier. Local use is free with no account and no usage cap. The paid tier is optional cloud inference, not a limit removal.
- How much does LM Studio cost?
- A free tier is published and no recurring figure is. The paid tiers appear on the vendor's own page without a monthly price beside them.
- Does LM Studio require a credit card to start?
- No card is needed for the free product this page describes. One is needed for Cloud, pay as you go, which is billed separately.
- Does LM Studio offer refunds?
- Not applicable to local use, which is free. Cloud credits are consumption-based.
- Does LM Studio have an app?
- Not a phone app. It installs on Windows, macOS and Linux, and there is no iOS or Android build.
- Does LM Studio have an API?
- Yes. An API is documented, so the same model can be called from your own code rather than only through the interface.
- Can LM Studio run locally?
- Yes. In its primary mode it runs models on your own hardware, so prompts and output stay on the machine and no vendor retention period applies to them. That holds until you use its optional hosted mode, which sends your prompts to a remote model provider under that provider's policy rather than this one. Cloud, pay as you go is a separate hosted service and none of that covers it.
- Does LM Studio train on my conversations?
- Nothing is transmitted in the local mode, so there is no user content for anyone to train on. That is a property of where the model runs rather than a promise that could be changed. Use the optional hosted mode and that provider's training terms apply to whatever you send it.
- What happens to my LM Studio data?
- There is no account and no vendor-held copy, so deleting means removing the files on your own device. Nothing has to be requested from anybody.
- Is LM Studio actually uncensored?
- ClickTug records no documented output filter, which is a different claim from having no rules: its terms still prohibit a specific list of conduct, set out above.
- What company owns LM Studio, and where?
- Element Labs, Inc., United States. Publishing an identifiable counterparty is worth noting: several tools in this directory name no company at all.
LM Studio FAQ
Is LM Studio free?
Yes, and the free version is the actual product rather than a trial. Running open-weight models on your own machine is unlimited, needs no account and costs nothing. LM Studio separately sells optional hosted inference billed per token, for models you would rather not run yourself, but nothing about the local application is gated behind it.
Is LM Studio safe?
Safe in which sense matters. On data, local use is the strongest position in this directory: models run on your hardware, LM Studio states no data ever leaves the device, voice transcription is processed on the machine, and no account is needed, so use can be genuinely anonymous. There is no retention policy to weigh because there is no server. The operator is Element Labs, Inc. of the United States. Two caveats belong in the same answer. The application applies no filter, so refusal behaviour is a property of the model you download. And an optional hosted mode plus the paid cloud tier are a separate hosted service under their own terms, where prompts do leave your machine.
Does LM Studio send my conversations anywhere?
Not when you use it locally. Models run on your hardware and LM Studio states that no data ever leaves your device, including voice transcription. There is no server-side chat history because there is no server involved. If you opt into the paid cloud inference, prompts do go to a hosted service, which LM Studio describes as United States based with zero data retention.
Is LM Studio uncensored?
The application applies no filter of its own, so refusal behaviour comes entirely from whichever model you download. Pick a permissive open-weight model and it will be permissive; pick a heavily aligned one and it will refuse. That is a real difference from a hosted service, where the filter belongs to the provider and can change without notice.
What hardware does LM Studio need?
More than a browser and less than you might fear. Smaller models run acceptably on a modern laptop with enough memory, while larger ones want a machine with a capable GPU or an Apple Silicon Mac with plenty of unified memory. The practical constraint is memory rather than raw speed, because a model has to fit before it can run.
Why does ClickTug list LM Studio without a minor safety policy?
Because it never receives user content. A local runtime has no upload to moderate, no account to age-check and no service to police, so there is nothing for such a policy to govern. That exemption applies only to tools that run on your own hardware, is recorded explicitly in our data, and does not extend to any hosted service including LM Studio's own optional cloud tier.
How we checked LM Studio
- Checked
- 8 August 2026
- Documents read
- 3
- Documentation gaps
- 3
What we did
- Read every policy document linked from this page, in full
- Loaded the pricing page, and re-loaded it in a rendering browser where it was empty
- Checked the homepage against the pricing and policy pages for disagreement
- Ran the vendor's own homepage through ClickTug's excluded-terms gate
- Searched for reporting, incident-database entries and independent testing
- Captured the interface and reviewed the capture by eye
What we have not done
Everything below needs an account, and ClickTug does not hold one here. These are recorded as not tested rather than guessed at, which is why they are absent from the tables above rather than marked "no".
- Signup completed
- Email verification required
- Card required at signup
- First response generated
- Long-term memory
- Voice feature
- Character creation
- Delete-account control
- Chat export
- Cancellation flow
Verified ClickTug observed this directly on the vendor's own pages.Documented The vendor states this in a document we have linked.Not documented We read the document that would carry this and it is absent.Not tested Checking this needs an account. We have not held one, so we have not checked.
Sources and identity
ClickTug last checked these facts on 8 Aug 2026 against 3 vendor sources. Pricing and data policies in this category change within weeks, so treat the date as part of the fact.
Product identity
| Official name | LM Studio | Verified |
|---|---|---|
| Operator | Element Labs, Inc. | Documented |
| Jurisdiction | United States | Documented |
| Website | lmstudio.ai | Verified |
| Also known as | LMStudio, lmstudio.ai | Verified |
Something wrong or out of date? Report this listing. If this listing concerns you personally, use our removal request form.
Ready to try LM Studio?
Checked on 8 Aug 2026 against 3 vendor sources.