Question 1

Is there an OpenAI-compatible API with no logging?

Accepted Answer

Yes. LibertAI exposes an OpenAI-compatible inference API at https://api.libertai.io/v1 that does not log your prompts and never trains on your data. It runs open-source models on Aleph Cloud's decentralized infrastructure, so you get the same request shape your code already uses without the data retention.

Question 2

How do I switch from the OpenAI API?

Accepted Answer

Because the API is OpenAI-compatible, switching is two lines of code. In any OpenAI SDK (Python or JavaScript) or framework like LangChain, set base_url to https://api.libertai.io/v1 and swap in a key from console.libertai.io. Your existing chat, completion, and embedding calls keep working with no other changes.

Question 3

Which open-source models can replace GPT-4 class models?

Accepted Answer

LibertAI serves open-weight models including Qwen3.6-35B-A3B and Qwen3.6-27B for long-context, tool use, vision, and reasoning, Gemma 4 31B with a 262K context window and multimodal support, DeepSeek V4 Flash for fast reasoning, GLM-4.7, Qwen3 Coder, and Qwen3.5 122B. There are also image generation, embeddings, text-to-speech, and a private web search API.

Question 4

Can I pay for an AI API with crypto?

Accepted Answer

Yes. You can fund your LibertAI account with crypto such as ETH and USDC, or with a credit or debit card. Pricing is pay-per-token, pay-as-you-go, and holding the LTAI token gives you a 20% discount on usage.

Question 5

What makes inference confidential?

Accepted Answer

TEE-served models, like Hermes 3 8B, run inside Trusted Execution Environments. A TEE isolates the computation so the host cannot read what is being processed, which makes the confidentiality verifiable rather than a promise. Combined with no logging and no training on your prompts, your inference stays private.