Can I run GLM 4.7 Flash Heretic without a GPU?

Yes, online in LU Cloud, where it is in the Uncensored group. It is also in the local catalogue of the app: the smallest build (IQ2_M) is 10 GB and fits a 12 GB card, and larger builds are 19, 25 and 32 GB.

Is Gemma 4 Uncensored available as a local download?

Not under that name. The local catalogue has other Gemma 4 builds, for example Gemma 4 12B Heretic at 7.4 GB. Aion 3.5 and Abliterated Large V2 have no local entry in the app's catalogue.

Do the new models carry the No Filter mark?

No. No Filter is the measured mark, and the ten models were not part of that measurement. Their rows carry the word Uncensored instead, and a row shows one of the two marks at most.

New Uncensored Chat Models in LU Cloud: GLM 4.7 Flash Heretic, Gemma 4 Uncensored and More, Online Without a GPU

Q: What is the Uncensored (Beta) group in LU Cloud?

Since 3.0.6 the Cloud model picker opens with a group of ten chat models whose makers release them to answer without refusing. LU Cloud serves them as published and never swaps one for another model. The group is in beta: the models are new, and how they hold up when many people use them at once is not measured yet.

Q: What does Limited capacity mean?

Some of the ten models show it in grey after their names. The tooltip reads: "All customers share a small number of requests a minute on this model. At busy times it asks you to wait a few seconds." It is about room, not about how the model answers.

Since Locally Uncensored 3.0.6, the Cloud model picker opens with a new group of ten chat models, marked Uncensored (Beta). Their makers release them to answer without refusing. The release notes name four of them: Gemma 4 Uncensored, GLM 4.7 Flash Heretic, Aion 3.5 and Abliterated Large V2.

This post covers what the group is, how it differs from the No Filter mark, what Limited capacity means, and how to use these models online without a graphics card of your own.

Beta, and it can change

The group is in beta. The models are new in LU Cloud, and how they hold up when many people use them at once is not measured yet. A model shows up once the cloud catalogue lists it, so your picker can differ from this post.

What the Uncensored Group Is

In Cloud mode the model picker shows the group under its own heading, Uncensored, with a Beta mark next to it. The models come from makers who publish them to answer without refusing. LU Cloud serves them as published and never swaps one for another model. Each row carries the word Uncensored where other rows show their mark, with the tooltip "Released by its makers to answer without refusing, and never swapped for another model."

We name four models because the release notes name them. The other six appear in your picker as the catalogue lists them.

The Four Named Models, Local or Online

This is what the local catalogue of LU 3.0.6 holds for the same names:

Model	Online in LU Cloud	On your own machine
GLM 4.7 Flash Heretic	Yes, in the Uncensored group	Yes. The smallest build (IQ2_M) is 10 GB and fits a 12 GB card. Larger builds are 19, 25 and 32 GB
Gemma 4 Uncensored	Yes, in the Uncensored group	No local entry under this name. The local catalogue has other Gemma 4 builds, for example Gemma 4 12B Heretic at 7.4 GB
Aion 3.5	Yes, in the Uncensored group	No local entry in the app's catalogue
Abliterated Large V2	Yes, in the Uncensored group	No local entry in the app's catalogue

So GLM 4.7 Flash Heretic runs both ways. Locally it runs on your card and nothing leaves your machine, but it wants at least a 12 GB card for the smallest build. Online it needs no download and no graphics card. For the other three, the cloud is the way to use them inside LU.

What Limited Capacity Means

Some of the ten models show Limited capacity in grey after their names. Hovering the note reads: "All customers share a small number of requests a minute on this model. At busy times it asks you to wait a few seconds."

So the note is about room, not about how the model answers. If you need steady speed for a long session, pick a model without it.

Uncensored Group and No Filter Mark

No Filter is the measured mark. Before 3.0.6 it was called No refusals. The name changed and the meaning did not. Its tooltip reads "Measured: this model answers without refusing." This is the sentence on the LU Cloud page:

19 of the 46 cloud chat models we measured answer in full without refusing, and only those carry the No Filter mark.

The ten models of the Uncensored group were not part of that measurement. Their mark rests on what their makers release them for, not on our test. A row shows one of the two marks at most.

Neither mark promises an answer to every prompt. The LU Cloud page says it for all hosted models: models can make mistakes or refuse requests. We add no content filter of our own to text.

How to Use Them Online

In the desktop app

Sign in under Settings, General, LU Cloud Account. An account is free
Click Cloud in the header next to the model picker. It turns into "Switch to Cloud?", and a second click within six seconds moves the whole app to Cloud
Open the model picker. The Uncensored group is there under its own heading. Pick a model and chat as usual

Switching back to Local brings your own models back. What Cloud use costs is on the pricing page. The handbook chapter on LU Cloud explains the account, the switch and the marks step by step.

In a browser

LU Cloud also runs in a browser with the same account, on any computer, including one where the desktop app does not run. The LU Cloud page sets out what the hosted side can and cannot do, and Uncensored AI chat online covers the chat in the browser.

What We Have Not Measured

We have not measured how the ten models hold up under load, and we have not run them through the test behind the No Filter mark. They are new in LU Cloud and can still change.

If you would rather keep everything on your own machine, the comparison of local AI apps and the chat chapter are the place to start.

Locally Uncensored is AGPL-3.0 licensed and built by PurpleDoubleD. Corrections welcome on GitHub Discussions or in the Discord.

In the handbook

LU Cloud, the account, the Cloud switch and the two marks in the picker
Chat, the model picker, the context window and thinking mode
First start, from the installer to your first local model