Frequently Asked Questions
Getting started
Section titled “Getting started”Gnoppix AI provides a unified API to access various LLMs. Please understand that we do not provide free hosted AI; instead, our platform is designed to run locally on your own hardware. Once installed, you have completely free, unlimited access to your local AI 24/7/365.
Sometimes, we test new models and make them available to our members for free. However, Gnoppix is not a charity providing free AI to everyone—we pay for all our hardware and operational costs ourselves, which is why free access is reserved exclusively for our members. Becoming a member is simple: if you contribute to the Gnoppix project in any way, you automatically join. The principle is simple: you help us, and we help you.
Pricing and Fees
Section titled “Pricing and Fees”Free for all members.
Is there a fee for using my own provider keys (BYOK)?
Section titled “Is there a fee for using my own provider keys (BYOK)?”No, you can use your own provider, Gnoppix API keys works on all Applications.
Models and Providers
Section titled “Models and Providers”Gnoppix offers specialized AI models tailored for privacy and industry-specific workflows, such as finance. By focusing on targeted fine-tuning, our models deliver more accurate domain-specific results than generic closed-source models.
Built to run locally, they stay 100% under your control. Our uncensored models answer every prompt directly and completely—without preachy disclaimers, patronizing warnings, or artificial restrictions.
Learn more at https://gnoppix.org/ai/
Check out the forum or Discord—if we receive a new quota, you’ll be informed there. Please note there may be a delay. If you’re interested in helping us maintain new models and keep them up to date, please contact us!
Again, our focus is not on building a business—it’s about providing you with your own local AI.
What LLM models does Gnoppix AI support?
Section titled “What LLM models does Gnoppix AI support?”For a complete list of models you can visit the list through the models api.
How frequently are new models added?
Section titled “How frequently are new models added?”We work on adding models as quickly as we can. We often have partnerships with the labs releasing models and can release models as soon as they are available. If there is a model missing that you’d like Gnoppix AI to support, feel free drop us a message Again if you would help us (*and the others) everyone is welcome to help.
What are model variants
Section titled “What are model variants”Some contingent models are mostly general-purpose. We don’t just distribute models—we design them. Typically, our models range from 0.9B to 550B parameters, running on everything from mobile devices to powerful local hardware.
I am an inference provider, how can I get listed on Gnoppix AI?
Section titled “I am an inference provider, how can I get listed on Gnoppix AI?”If you would like to contact us, the best place to reach us is over our Ticket System.
Be aware you need to offer your models for testing/free. We do not sell/resell.
Why are there free hosted AI providers? They’re new and testing their infrastructure. You’re the testers, and in return, you get free access. You can use the data to train your own models or develop new ones. Nothing in this world is truly free; you always pay in one way or another.
What is the expected latency/response time for different models?
Section titled “What is the expected latency/response time for different models?”We have a world-wide enabled smart routing with servers all over the world, which balances and optimizes for low latency. Nevertheless, if you contact a provider by yourself, there is also latency included in your queries, and this can vary significantly between different models and providers.
The expected latency/response time for different LLM models is highly variable and depends on numerous factors, including:
-
Model Size and Complexity: Larger, more complex models (e.g., Opus, GPT-5.6) generally have higher latency due to the increased computational resources required for inference. Smaller, more optimized models (e.g., Flash versions, mini models, Groq-hosted Llama) tend to be much faster.
-
Prompt Length (Input Tokens): Longer input prompts require more processing time, increasing the “Time to First Token” (TTFT).
-
Response Length (Output Tokens): LLMs generate responses token by token. The total response time is directly proportional to the number of output tokens.
-
Hardware and Infrastructure: The type and configuration of GPUs/TPUs, memory bandwidth, and the efficiency of the inference pipeline (e.g., batching, simultaneous inference) significantly impact speed.
-
Network Conditions and Geographic Location: The physical distance between your application and the LLM provider’s data center, as well as general internet congestion, adds to latency.
-
Provider Optimization: Different providers optimize their serving infrastructure differently. Some focus on raw speed, others on cost-efficiency, and some on maximizing throughput.
-
“Cold Starts”: If a model instance needs to be spun up from an inactive state, the first request might experience higher latency.
-
Specific Task: Latency can vary by use case (e.g., coding, summarization, creative writing), as different tasks may stress different parts of the model or require different processing steps.
General Observations (based on recent benchmarks):
Fastest (Lower Latency, Higher Tokens/Second):
- Models specifically designed for speed, like Gemini Flash versions, Groq-hosted Llama models, and certain Mistral or Llama 3.1 8B Instruct variants, often exhibit latencies in the range of 0.3 to 1.0 seconds for Time to First Token and high throughputs (e.g., 50+ to over 200 tokens/second). These are ideal for real-time interactive applications. As example Agents.
Balanced Performance:
-Models like OpenAI’s GPT, and Anthropic’s Claude Sonnet often strike a balance between capability and speed, with TTFT generally in the 0.4 to 2.0 second range and good token generation rates (e.g., 60-80 tokens/second).
Higher Latency (for more complex tasks or larger models):
-
Larger, more powerful models such as Anthropic’s Claude Opus, GPT-5.x, or complex Gemini Pro versions can have higher latencies, sometimes in the 2.0 to 5.0+ second range for Time to First Token, particularly for longer inputs or more involved reasoning. These are often chosen when quality and reasoning capabilities are prioritized over instantaneous response.
-
Some “preview” or very large models might also show higher latencies due to their experimental nature or sheer size (e.g., GPT-5 Preview).
How does model fallback work if a provider is unavailable?
Section titled “How does model fallback work if a provider is unavailable?”If a provider returns an error Gnoppix AI will automatically fall back to the next provider. This happens transparently to the user and allows production apps to be much more resilient.
Privacy and Data Logging
Section titled “Privacy and Data Logging”Please see our Terms of Service and Privacy Policy. We take your privacy and security very seriouse. We always uggesst use Open Source LLM before Closed Source.
What data is logged during API use?
Section titled “What data is logged during API use?”We log basic request metadata (timestamps, model used, token counts). Prompt and completion are not logged by default. We do zero logging on our site of your prompts/completions, even if an error occurs. Be aware we have a IDS (AI based of course to detect missuse, brake=ins) If you’re blocked for some reason contact us. or wait for 30 Days the system automatically unblock you. We collect the ‘bad guys’ in our RBL https://rbl.gnoppix.org
What data is logged during Chatroom use?
Section titled “What data is logged during Chatroom use?”All conversations in the chatroom are stored locally on our Server. Conversations will sync across devices. It is possible to exporti/export and delete conversations using the settings menu in the chatroom.
What third-party sharing occurs?
Section titled “What third-party sharing occurs?”Gnoppix AI is a proxy that sends your requests to the GPU provider for it to be completed (which are our own GPU Servers)
We do not guarantee that your prompts will not be stored or further processed by 3rd.Party providers.
Credit and Billing Systems
Section titled “Credit and Billing Systems”What purchase options exist?
Section titled “What purchase options exist?”none, use local AI, it is free :)
How to monitor usage?’>
Section titled “How to monitor usage?’>”The Activity page allows users to view their historic usage and filter the usage by model, provider and api key.
What free tier options exist?
Section titled “What free tier options exist?”All Gnoppix Member receive a very small free allowance to be able to test out our own generated Models.
There are many free models available on Gnoppix AI, it is important to note that these models have low shared rate limits 1000 requests per day total and are usually not suitable for production use.
How does Gnoppix AI make money?
Section titled “How does Gnoppix AI make money?”We don’t. We have our own models. In some cases, companies give us free credits for testing their models.
Account Management
Section titled “Account Management”How can I delete my account?
Section titled “How can I delete my account?”Go to the Settings page and click Manage Account. In the modal that opens, select the Security tab. You’ll find an option there to delete your account.
How can I contact support?
Section titled “How can I contact support?”The best way to reach us is to join our Discord and ping us in the #help forum. Alternative you can send a message Ticket System