Understanding Uncensored vs. Abliterated Models
The primary distinction lies in how the model handles constraints. Standard models use Reinforcement Learning from Human Feedback (RLHF) to refuse topics. Abliterated models remove the refusal mechanism but often retain the underlying biases, making them appear censored in practice.
When you need a reliable uncensored LLM for roleplay or creative writing, you want a model that answers without hesitation. Our dedicated inference server serves one such model tuned for lawful adult use. It does not block political, sexual, or controversial topics. You avoid the unpredictability of abliterated variants that may still refuse valid prompts.
API Compatibility: The OpenAI Standard
We follow the OpenAI API standard for maximum compatibility. You can use any existing OpenAI SDK or client library with our endpoint. Just update the base URL and API key in your configuration.
The endpoint is POST /v1/chat/completions. You send a model ID of "uncensored" and receive text responses. We do not offer embeddings, image generation, or audio processing. This focused approach ensures high availability and predictable latency. Use GET /v1/models to verify available endpoints.
This standardization means you can switch between vendors if needed, but our privacy guarantees and crypto billing make us a strong alternative. Check the official documentation for specific parameter details.
Token Economics and Cost Management
Our billing is transparent and prepaid. You pay $0.15 per 1 million input tokens and $0.50 per 1 million output tokens. Errors and refusals do not charge you. This protects your budget during testing phases.
You top up with crypto only. Accepted currencies are USDT (TRC20) or USDC (Base). Minimum top-ups start at $20. Bonuses apply automatically: +5% for $50, +10% for $100, and +15% for $200. Credit never expires, so you can use it whenever needed.
No monthly subscriptions. No hidden fees. Just pay for what you use. This model suits developers who want predictable costs without recurring charges. Track your usage via the dashboard to manage your budget effectively.
Context Window and Output Limits
The context window supports up to 100,000 tokens. This includes both your prompt and the completion. You can send long documents or maintain extensive conversation histories. This is ideal for complex automation tasks.
However, the maximum output per request is 16,000 tokens. If you do not set max_tokens, it defaults to 2,048 tokens. Plan your prompts accordingly to avoid truncation. Adjust the max_tokens parameter based on your specific use case.
This balance allows for deep context understanding while keeping response times manageable. Use this capacity to feed large datasets or maintain state in chatbots. Monitor your token usage to optimize costs and performance.
Streaming and Real-Time Generation
We support streaming via Server-Sent Events (SSE). This allows you to receive tokens as they are generated. You get token usage data in the last chunk. This is crucial for real-time applications like chatbots.
Streaming reduces perceived latency for end-users. They see text appearing in real-time. Use the official SDKs to handle streaming easily. Our API ensures smooth data flow without interruptions.
Combine streaming with function calling for dynamic interactions. You can parse tool outputs on the fly. This flexibility makes our API suitable for complex workflows. Experiment with different streaming configurations to find the best user experience.
Rate Limits and Concurrency
Our API enforces clear limits to ensure stability. You can make 300 requests per minute per key. Concurrency is limited to 8 simultaneous requests per key. The maximum request body size is 8 MB.
These limits prevent abuse and maintain performance. If you exceed the rate limit, you will receive an error. Monitor your request volume to stay within bounds. One active key per account ensures accountability.
Replace the old key if needed. The new key starts fresh. This simplicity helps developers manage their API keys efficiently. Use rate limiting in your client code to handle retries gracefully. Plan your concurrency based on your expected load.
Privacy: Prompt Usage and Data Retention
Privacy is a core feature. We do not use your prompts for training. Your data remains yours. This is different from many free or cheap APIs that monetize your usage.
Account creation requires only an email. No phone number needed. You can sign in with Google or use email and password. This minimizes your digital footprint. Crypto payments add another layer of anonymity.
Your prompts are not stored indefinitely. We focus on serving your requests, not harvesting your data. This trust is essential for sensitive applications. Use our API for confidential projects without worrying about data leakage. Check the privacy policy for more details.
Content Limits and Refusal Logic
Our model is uncensored but has one hard limit. It refuses sexual content involving minors. This is the only strict refusal. All other lawful adult content is allowed.
This clarity helps you build reliable applications. You do not need to handle unexpected refusals for standard NSFW content. The model answers directly. This consistency is vital for automation pipelines.
You can generate creative stories, roleplay, or technical content without filters. Use the API for diverse use cases. The lack of bias means you get raw, unfiltered outputs. This is perfect for users who want full control over their content.
Integration Steps for Developers
Start by signing up on the "Get API key" page. You can use Google or email. Your key is shown immediately. No phone number required.
Top up your credit with crypto. Use USDT or USDC. The minimum is $20. Bonuses apply automatically. Start making requests to POST /v1/chat/completions. Set the model to "uncensored".
Use the official SDKs for easy integration. Handle streaming for real-time responses. Monitor your usage and limits. Replace keys if needed. This straightforward process gets you up and running quickly. Focus on building your application, not managing infrastructure.
