Is the API actually uncensored?
Yes, the model is tuned to answer without content refusals for lawful adult use, fiction, or controversial topics. Unlike corporate models that hedge or refuse based on broad safety guidelines, this open-weight model prioritizes direct answers. It does not apply the same rigid filters that block creative or mature themes. However, one hard limit always applies: sexual content involving minors is refused. You get raw output, but you still need to handle that specific edge case in your own application logic. This makes it ideal for NSFW chatbots or unrestricted story generation where standard models cut off mid-sentence.
Check the Context Window
The context window supports 100,000 tokens for both prompt and completion combined. This allows you to pass large documents or long conversation histories without losing earlier data. When designing your system, ensure you truncate or summarize older messages if you approach this limit. The model retains coherence across long contexts, which is critical for maintaining character voice in roleplay scenarios. If you send a prompt that exceeds the available space, the API will return an error. Plan your token budget carefully to avoid unexpected truncation during peak usage.
Verify Output Limits
Each request can generate up to 16,000 tokens of output. If you do not explicitly set the max_tokens parameter, the default is only 2,048 tokens. This distinction matters for long-form content generation or detailed code outputs. Without setting the limit, you might get truncated responses that look like errors. Always define max_tokens in your API calls to ensure you receive the full answer. This flexibility supports everything from short replies to lengthy essays without needing multiple round-trips.
Streaming and Token Counts
Streaming is supported via Server-Sent Events (SSE), allowing real-time token delivery to your frontend. The final chunk of the stream includes the total token usage for that request. This feature lets you track costs accurately without making a separate call to a usage endpoint. You can display tokens as they generate, improving the user experience for slow responses. Combine this with your billing logic to charge users based on actual consumption. It is a standard feature that simplifies integration with existing OpenAI-compatible clients.
Function Calling Support
Function calling is fully supported, allowing you to define tools and specify tool_choice. This enables structured data extraction or automated actions based on user input. You can send a list of function definitions in the request, and the model will return the appropriate function call in the response. This is essential for building agents that interact with external APIs or databases. Ensure your client library supports the tools parameter for seamless integration. The model handles multiple functions effectively, maintaining context across tool use.
JSON Mode Reliability
JSON mode is available via the response_format parameter set to json_object. This forces the model to output valid JSON, which is crucial for parsing data in your application. While not perfect, it significantly reduces the need for regex post-processing compared to standard text outputs. Use this for structured data extraction or when your frontend expects a specific schema. It works well with function calling and streaming. Test your specific prompts to ensure the JSON structure meets your requirements before scaling up.
Pricing Transparency
Pricing is straightforward: $0.15 per 1M input tokens and $0.50 per 1M output tokens. You pay only for what you use, with prepaid credit that never expires. Errors and refusals do not consume credits, so you do not pay for failed requests. You can top up with USDT on TRC20 or USDC on Base, with amounts ranging from $20 to $500. Bonus credit applies at higher tiers: +5% from $50, +10% from $100, and +15% from $200. No subscription fees or hidden costs mean you control your budget exactly.
Privacy and Data Usage
Your prompts are not used for training the model, ensuring your data remains private. The API is hosted on our own servers, so you do not rely on third-party aggregators. You can sign up with just an email address, keeping your identity minimal. Data retention policies are clear, and you can manage your account securely. For developers concerned about proprietary data or sensitive conversations, this model offers a reliable alternative to public APIs that may use data for improvement. It is a clean slate for your projects.
Rate Limits and Concurrency
You are limited to 300 requests per minute and 8 concurrent requests per API key. This ensures fair usage and prevents any single user from overwhelming the servers. If you need higher throughput, you can generate additional keys, though each account is limited to one active key at a time. The request body size is capped at 8 MB, which is sufficient for most text-based interactions. Monitor your usage to stay within these limits and avoid 429 errors during peak times. It is a balanced setup for both small and large-scale applications.
