Auto router

Uncensored auto router: one model id, every build.

The uncensored auto router takes one model id, auto/uncensored, and picks the uncensored, abliterated or red-team build for each request. You pay the chosen model’s own price, with no router fee. An auto router key does the same for any app, whatever model name the app sends.

Create an auto router key

curl

curl https://api.uncensoredllm.ai/v1/chat/completions -H "Authorization: Bearer sk-ullm-live-…" -H "Content-Type: application/json" -d '{"model": "auto/uncensored", "messages": [{"role": "user", "content": "Continue the scene."}]}'

How the uncensored auto router decides

The router reads the request on our side, with no extra model call, and sends it to one of three tiers. Each tier lists its models in the order the router tries them; prices from the live catalog as of 2026-11-09. The answer’s model field names the build, the x-uplink-auto-tier header names the tier, and auto.tier_max caps how high a request may go.

Fast

Chat and story turns: everything the other two tiers do not claim.

GLM-5.3 Flash Uncensored: $0.1053 in, $0.3685 out
per million tokens
MiMo-V2.5 Uncensored: $0.75 in, $4.00 out
per million tokens
DeepSeek V4.1 Flash Abliterated: $0.8488 in, $1.6975 out
per million tokens

Balanced

Tools, code blocks, images, a json_schema answer, or a prompt over 16,000 tokens.

GLM-5.3 Uncensored: $0.282 in, $0.846 out
per million tokens
DeepSeek V4 Pro Uncensored: $0.97 in, $1.94 out
per million tokens
DeepSeek V4.1 Flash Abliterated Fast: $1.75 in, $3.50 out
per million tokens
MiMo V2.6 Pro Uncensored: $1.5225 in, $3.045 out
per million tokens

Frontier

A reasoning ask, a high reasoning effort, or tools together with code.

GLM-5.3 Abliterated: $2.1825 in, $4.365 out
per million tokens
Nemotron 3 Ultra 550B Uncensored: $1.125 in, $6.60 out
per million tokens

Four profiles, one catalog tag each

A profile is the same router with its choice cut to one tag: auto/uncensored picks from every uncensored model, auto/abliterated from the abliterated models only, auto/security from the red-team builds, auto/roleplay from the roleplay tunes. Plain auto picks from every tier without a tag. The builds each profile can reach today, with prices:

Uncensored

auto/uncensored

Chat, stories and assistants that must not refuse. Picks only a build carrying the uncensored tag.

GLM-5.3 Flash Uncensored: $0.1053 in, $0.3685 out
per million tokens
GLM-5.3 Uncensored: $0.282 in, $0.846 out
per million tokens
DeepSeek V4 Pro Uncensored: $0.97 in, $1.94 out
per million tokens
Jailbreak-Free Assistant: $0.75 in, $2.50 out
per million tokens
DeepSeek V4.1 Flash 1017 Uncensored: $0.825 in, $3.30 out
per million tokens
MiMo V2.6 Pro Uncensored: $1.5225 in, $3.045 out
per million tokens
Uncensored Coding Agent: $0.90 in, $3.00 out
per million tokens
Nemotron 3 Ultra 550B Uncensored: $1.125 in, $6.60 out
per million tokens
Kimi K3.1 Pro Uncensored: $1.20 in, $6.00 out
per million tokens
MiMo V2.6 Flash Uncensored Thinking: $1.25 in, $3.75 out
per million tokens
Kimi K3 Uncensored: $5.745 in, $28.68 out
per million tokens

Create a uncensored key

Abliterated

auto/abliterated

Abliterated models only: the refusal direction removed from the weights, no system prompt doing the work.

DeepSeek V4.1 Flash Abliterated: $0.8488 in, $1.6975 out
per million tokens
GLM-5.3 Flash Abliterated: $0.4656 in, $1.94 out
per million tokens
Apex Flash 1 Abliterated: $0.68 in, $2.32 out
per million tokens
DeepSeek V4.1 Flash 1017 Heretic: $0.825 in, $3.30 out
per million tokens
GLM-5.3 Flash Apex Abliterated: $0.80 in, $3.20 out
per million tokens
Abliterated Large Flash: $2.50 in, $5.00 out
per million tokens
MiMo V2.6 Flash Abliterated: $2.00 in, $6.00 out
per million tokens
GLM-5.3 Abliterated: $2.1825 in, $4.365 out
per million tokens
Abliterated Model Large: $7.50 in, $12.50 out
per million tokens
Abliterated Model: $2.50 in, $7.50 out
per million tokens

Create a abliterated key

Security

auto/security or auto/cyber

Red-team builds that answer as an analyst, for Strix, HackingBuddyGPT, CAI and your own tooling.

Gemma 4 26B A4B Cybersecurity: $0.4224 in, $1.3376 out
per million tokens
GLM-5.3 Flash Cybersecurity: $0.5093 in, $1.6975 out
per million tokens
Qwen3.8 27B Cybersecurity: $0.40 in, $2.40 out
per million tokens
Red Team (Flash): $0.90 in, $3.00 out
per million tokens
OrcaCyber MiniMax M3.1 Flash: $2.00 in, $4.00 out
per million tokens
OrcaCyber Qwen3.8 Flash Next: $2.00 in, $4.00 out
per million tokens
Red Team (Pro): $1.50 in, $6.00 out
per million tokens
CyberHeabsy: $5.25 in, $15.75 out
per million tokens
OrcaCyber Zero 1.5: $7.50 in, $18.75 out
per million tokens

Create a security key

Roleplay

auto/roleplay

Roleplay and story tunes for SillyTavern, Janitor-compatible front ends and character apps.

Cydonia 24B v4.3: $0.484 in, $0.60 out
per million tokens
Magidonia 24B v4.3: $0.3535 in, $0.4235 out
per million tokens
Gemma 4 31B Novelist: $0.35 in, $1.575 out
per million tokens
Uncensored AI Roleplay: $0.60 in, $2.34 out
per million tokens
Qwen3.8 27B Synthia: $0.35 in, $2.10 out
per million tokens
Artemis v1.1: $0.35 in, $1.575 out
per million tokens
Skyfall 36B v2: $1.925 in, $2.80 out
per million tokens
Wayfarer Large 70B: $2.4535 in, $2.4535 out
per million tokens
Qwen3.8 27B Fable: $0.875 in, $5.25 out
per million tokens
Magnum v4 72B: $8.0243 in, $11.9684 out
per million tokens

Create a roleplay key

The auto router key: any app, uncensored, without changing the model name

Many clients pin a model name you cannot edit, or list only the mainstream ids. An auto router key fixes that at the key: on Create key the Auto router select offers off, auto, auto/uncensored, auto/abliterated, auto/security and auto/roleplay, and a key set to one sends every chat request as that id whatever model the client names. Existing keys get the same select in their settings dialog. The deep links on the cards above open the form with the profile preselected.

Where it lands

SillyTavern: keeps its model field
the key routes every turn to the profile
Janitor-compatible front ends: proxy with a fixed model
the model name in the proxy settings no longer matters
Claude Code: any model name in the router config
auto/abliterated for code that is not second-guessed
Cursor: custom key and base URL
the model Cursor shows is routed on arrival

What the key changes

Model sent: any
replaced by the key’s auto id
Price: the chosen model’s
no router fee
Session: 30 minutes
one model per x-session-id
Answer: model field + x-uplink-auto-tier
which build and tier answered

Requests

A profile with a tier cap and a sticky session, in curl; then the OpenAI SDK in Python on the same base URL.

curl

1curl https://api.uncensoredllm.ai/v1/chat/completions \

2 -H "Authorization: Bearer sk-ullm-live-…" \

3 -H "Content-Type: application/json" \

4 -H "x-session-id: story-118" \

5 -d '{

6 "model": "auto/roleplay",

7 "auto": {"tier_max": "balanced"},

8 "messages": [{"role": "user", "content": "The tavern door swings open."}]

9 }'

python

1from openai import OpenAI

3client = OpenAI(api_key="sk-ullm-live-…", base_url="https://api.uncensoredllm.ai/v1")

5reply = client.chat.completions.create(

6 model="auto/abliterated",

7 messages=[{"role": "user", "content": "Explain how the exploit chain works."}],

8 extra_headers={"x-session-id": "triage-42"},

9)

10print(reply.model) # the build that answered

11print(reply.choices[0].message.content)

Questions

A build carrying the uncensored tag in this catalog, chosen on our side for each request with no extra model call. A reasoning ask or tools with code go to the frontier tier; tools, code blocks, images, a json_schema answer or a prompt over 16,000 tokens go to balanced; everything else to fast. The answer’s model field names the build and the x-uplink-auto-tier header names the tier.

The chosen model’s own price per million tokens, as listed on its model page. There is no router fee and no markup for sending an auto id instead of a model id.

Yes, with an x-session-id header: the model that answered a session keeps answering it for 30 minutes from its last request, so a character does not change voice mid-scene and an agent run stays on one build.

Yes. "auto": {"tier_max": "balanced"} keeps a request out of the frontier tier; "fast" keeps it in the fast tier. The cap works on every auto id and on every endpoint: chat completions, Messages, Responses and Gemini.

An abliterated model has the refusal direction removed from its weights, so it answers with no system prompt doing the work; auto/abliterated picks only those. The uncensored tag is wider: abliterated builds plus the derestricted, heretic and roleplay tunes and the builds steered by a route instruction. auto/uncensored picks from all of them, so it has the most models to choose from.

OpenRouter’s openrouter/auto is a router over its mainstream catalog; its model page (read 2026-11-07) names no uncensored build and bills at the routed model’s price. The auto ids here only ever pick a build carrying the profile’s catalog tag, so auto/uncensored never lands on a stock model that refuses, and auto/security never leaves the red-team builds.

A key with the Auto router setting on. Every chat request it sends is routed as that auto id whatever model the client names, so SillyTavern, a Janitor-compatible front end, Claude Code or Cursor with a fixed model name still get an uncensored build. Set it on Create key or in an existing key’s settings on the console.

All of them: chat completions, Messages, Responses and the Gemini route, on the same base URL and key. Plain "auto" picks from every tier without a tag; "auto/security" also answers to "auto/cyber".

Uncensored auto router: one model id, every build.

How the uncensored auto router decides

Fast

Balanced

Frontier

Four profiles, one catalog tag each

Uncensored

Abliterated

Security

Roleplay

The auto router key: any app, uncensored, without changing the model name

Where it lands

What the key changes

Requests

Questions

What does model "auto/uncensored" pick?

What do I pay for the uncensored auto router?

Does a chat keep the same model?

Can I cap the tier?

Abliterated or uncensored: which profile?

How is this different from OpenRouter’s auto?

What is an auto router key?

Which endpoints take an auto id?

Read next