Open WebUI needs one connection to this API, and after that your key decides which uncensored builds show up in its picker. Below: what you need, the quickest route to a first reply, the connection fields one by one, a trimmed picker, a preset per build, and the context and cost of a chat. Field names were read from Open WebUI's docs and its v0.12.3 source on 25 October 2026.
New to builds with the refusals removed? Read AI with no restrictions first, then connect them here.
What you need
- A running Open WebUI where you are an admin. Only admins see the Connections settings.
- An API key from the console. Keys start with
sk-unr-and are displayed a single time, so copy the key before you close the dialog. Top up the balance from $10 by card, WeChat Pay or crypto. - The base URL
https://api.unrestrictable.ai/v1. No install is needed on the API side: Open WebUI's docs state that any backend that talks OpenAI Chat Completions can be connected, and each build in this catalog does.
No Open WebUI yet? The same builds answer in the browser at /chat, three free turns a day with no account, and over Telegram at /telegram. Come back here when you want them inside your own interface.
Every field name below was checked against the OpenAI-compatible connection doc and the source at tag v0.12.3, the newest release as of 25 October 2026 (it shipped 16 October).
The shortest path to a first reply
- Signed in as admin, open the connection settings. The doc's path is Settings > Admin > Connections.
- Find Manage OpenAI API Connections and click the plus icon, Add Connection.
- URL takes
https://api.unrestrictable.ai/v1; API Key takes the key. - Keep API Type at Chat Completions. That is the route the daily check exercises on every build, at
/v1/chat/completions, so it is the one known to work. - Click Verify Connection and wait for "Server connection verified". Only then click Save: per the doc, Save on its own runs no test.
- Start a chat. The model picker now holds whatever ids the key is allowed to call. Choose
glm-5.3-uncensoredand type something.
You are connected. Everything after this point is about living with it: fewer entries in the picker, a preset for each build, and a feel for context and cost.
The connection fields, explained
Add Connection shows more inputs than this API uses. The ones that matter are below; the rest can stay as Open WebUI sets them.
- URL
- https://api.unrestrictable.ai/v1Includes /v1; nothing after it
- Auth
- BearerAlready selected
- API Key
- sk-unr-live-…From the console; displayed a single time
- API Type
- Chat CompletionsAlready selected; do not switch to Responses
- Model IDs
- (empty, or the ids below)Blank shows every id the key may call
- Prefix ID
- (empty)For two connections whose ids collide
Prefix ID exists to tell apart two connections that expose the same id: it prepends a label in the picker and, according to the source, removes it again before sending. One connection to this API has no collisions, so leave the field blank.
Trim the model picker
The picker is built from /v1/models, which returns every id your key may call. By default that includes all 169 uncensored builds on the models page, and a list that long gets in the way. You can shorten it in two places, and the two do different things.
On the key, in the console
Open the key's settings in the console, enter the ids you want in Models as a comma-separated list, and click Set. /v1/models then returns only those ids, and Open WebUI's picker shrinks without any change in Open WebUI. Any other id is turned away before a model sees it, at no charge:
glm-5.3-uncensored, deepseek-v4-pro-uncensored, nemotron-3-ultra-550b-a55b-uncensoredHTTP 403{"error":{"message":"This key may not call kimi-k3-uncensored.","type":"permission_error","param":null,"code":"scope_denied"}}Use this one when other people share your Open WebUI. Whatever they type, a build missing from the field stays out of reach.
On the connection, in Open WebUI
The connection's Model IDs field trims the list on Open WebUI's side only. Enter an id, click the plus icon, repeat. Blank, it accepts the whole /models answer. This decides what is displayed, not what the key is permitted to call.
There is a third control at Admin Settings > Models. The Workspace models doc explains that Enabled takes a model away from every user, while Hide only drops it from the picker.
Four builds worth listing
The set below is four general chat builds. Prices and windows are whatever the catalog held at render time. The daily check column is read from the daily real-model check. That check sends one fixed prompt to each model every day and records whether the reply still matches the labelled model.
| Wire formats | |||||
|---|---|---|---|---|---|
| GLM-5.3 Uncensored | Uncensored | $0.282 | $0.846 | — | |
| DeepSeek V4 Pro Uncensored | Uncensored | $0.97 | $1.94 | — | |
| Nemotron 3 Ultra 550B Uncensored | Uncensored | $1.0913 | $6.402 | 262K |
| Model id | Daily check | Context window |
|---|---|---|
glm-5.3-uncensored | Steady since Oct 12 | not published |
deepseek-v4-pro-uncensored | Steady since Oct 12 | not published |
nemotron-3-ultra-550b-a55b-uncensored | Steady since Oct 27 | 262,144 tokens |
There are more GLM builds than glm-5.3-uncensored; the uncensored GLM page has the others. Two or three ids are enough to begin with. A later addition is one more entry in the key's Models field.
A Workspace preset for each build
A preset is a base model wrapped with a name, a system prompt and parameters of its own, and your users choose the preset rather than the bare id. The field names come from the Workspace models doc.
- Go to Workspace, open the Models tab and click Create.
- Enter a name and an ID, say "Long-form writer". Spaces are not allowed in the ID.
- Choose one of the ids above as Base Model (From).
- Fill System Prompt with a straightforward brief covering tone, length and format. These builds do not refuse, so you need no coaxing text.
- In Advanced Params, give max_tokens a value to cap reply length, and lower Temperature if the default runs too hot for you.
- Access decides who sees it: you alone, a group, or everyone.
Last, hide the raw base models under Admin Settings > Models so only presets show in the picker. Per the doc, a hidden base model keeps serving the presets that sit on it.
Context: what each turn carries
Each message from Open WebUI carries the whole conversation to that point. By turn 40 of the example chat below, one request holds roughly 21,900 tokens. The window listed for each build in the table above is the hard limit on that request.
Advanced Params also offers num_ctx (Ollama). The bracket tells you its scope: it is Ollama's context-size setting. In the v0.12.3 source, advanced params for a non-Ollama model are copied into the request body unchanged, and num_ctx is not part of the OpenAI format, so on this connection it changes nothing. max_tokens, by contrast, is passed through to the API.
Once a chat runs long, open a fresh one or switch to a build whose published window is bigger. Among these four, nemotron-3-ultra-550b-a55b-uncensored publishes the largest.
The bill for a chat
Assume a system prompt of 300 tokens, user messages of 150 and replies of 400. Over 10 turns the input adds up to 29,250 tokens. Over 40 turns it is 447,000, since each turn repeats all the earlier ones.
| Model | 10-turn chat | 40-turn chat |
|---|---|---|
| GLM-5.3 Uncensored | $0.01 | $0.14 |
| DeepSeek V4 Pro Uncensored | $0.04 | $0.46 |
| Nemotron 3 Ultra 550B Uncensored | $0.06 | $0.59 |
Multiplying the turns by four multiplies the input by roughly 15. Open a new chat for each topic and the bill stays level. Per-token rates for every build are on the pricing page.
Test the key from a terminal
Verify Connection hits /models, the same endpoint as the request below. The call is free, so run it before you blame Open WebUI for a failure.
curl https://api.unrestrictable.ai/v1/models \ -H "Authorization: Bearer sk-unr-live-…"A 200 carrying a list of ids confirms the key. With the key's Models field set, only those ids are in the list. A 401 tells you which way the key is wrong:
HTTP 401{"error":{"message":"No API key was sent.","type":"authentication_error","param":null,"code":"invalid_api_key"}}HTTP 401{"error":{"message":"The API key is not valid.","type":"authentication_error","param":null,"code":"invalid_api_key"}}When Open WebUI does not connect
401 on Verify Connection
API Key is blank, or holds a key that is mistyped, truncated or revoked. Make a new one in the console and paste the whole string.
404 model_not_found
An entry in Model IDs, or the base model of a preset, does not match any catalog id. Take ids from the table above or the models page rather than typing them:
HTTP 404{"error":{"message":"No such model: glm-5.3-uncensord.","type":"not_found_error","param":null,"code":"model_not_found"}}402 insufficient_balance
Nothing is left on the prepaid balance. Top up in the console. A request that gets a 402 costs nothing.
HTTP 402{"error":{"message":"The prepaid balance does not cover this request. Top up at … to continue.","type":"insufficient_quota","param":null,"code":"insufficient_balance"}}403 scope_denied
That id is missing from the key's Models field. Either add it there or remove it from the connection's Model IDs so it cannot be picked.
An empty or slow picker
Make sure the connection is switched on and that every entry in Model IDs is a real id. For a model list that times out, the doc points to AIOHTTP_CLIENT_TIMEOUT_MODEL_LIST, which defaults to 10 seconds. The docs page lists the base URL, the key and the model id for any other client.
Questions
Under Settings > Admin > Connections, add a connection with https://api.unrestrictable.ai/v1 as the URL and your key, click Verify Connection, then Save. The picker fills with every id the key may call; set the key's Models field in the console to shorten it.
Yes. Open WebUI documents support for any backend that speaks the OpenAI Chat Completions protocol, and every uncensored build in this catalog speaks it. See the uncensored model list.
https://api.unrestrictable.ai/v1, with /v1 at the end. Open WebUI appends /models, /chat/completions and the other paths itself.
In the connection's API Key field, leaving Auth on Bearer. Open WebUI then sends it as a Bearer token with each request.
Yes. Open WebUI's API Endpoints reference lists POST /api/chat/completions, authenticated with a Bearer token from Settings > Account (read 25 October 2026). That API is for scripts that talk to your Open WebUI; the connection in this guide is the model backend it calls.
Chat Completions, which is the default. The daily model check runs every build through /v1/chat/completions. See the daily real-model check.
Every model above is on the uncensored model list with its live price per million tokens. Per-token pricing sets the whole catalog side by side, and the family directory groups the builds by base model.