# Uncensored LLM API

**NEAR AI Cloud is an uncensored LLM API. It serves abliterated open-weight models, with their refusal behavior removed, through an OpenAI-compatible endpoint. The models run inside secure enclaves (TEEs), so neither NEAR AI nor the cloud provider can read your prompts. Every account can use them, with no application.**

## Key points

- Models: abliterated DeepSeek V4.1 Flash and Qwen 3.8 27B. Live context windows and prices are at [near.ai/models.md](https://near.ai/models.md) and on each model page.
- OpenAI-compatible: base URL `https://cloud-api.near.ai/v1`, chat completions, your NEAR AI API key.
- Private: the models run in trusted execution environments on Intel and NVIDIA confidential computing. You can verify it with a hardware attestation.
- Open to every account. No application, no identity checks, pay per token with no minimums.
- Use is governed by the [Acceptable Use Policy](https://near.ai/acceptable-use-policy).

## Why private matters for uncensored work

You reach for an uncensored model when the work is sensitive: an exploit for your own system, an unpatched vulnerability, a malware sample, a hostile bug report. On a typical hosted API, those prompts are readable on the provider's servers, and "we don't log" is a promise you cannot check. NEAR AI Cloud runs its abliterated models inside TEEs, so NEAR AI and the cloud provider cannot see your prompts or the answers. See [What is private AI inference?](https://near.ai/private-ai-inference)

## Quickstart

1. Create a NEAR AI Cloud account at [cloud.near.ai](https://cloud.near.ai).
2. Create an API key and export it as `NEARAI_API_KEY`.
3. Set the base URL to `https://cloud-api.near.ai/v1` and call a model ID.

```bash
curl https://cloud-api.near.ai/v1/chat/completions \
  -H "Authorization: Bearer $NEARAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-ai/DeepSeek-V4.1-Flash-1017-abliterated",
    "messages": [
      {"role": "user", "content": "Explain how this SQL injection payload works."}
    ]
  }'
```

```python
from openai import OpenAI
import os

client = OpenAI(
    base_url="https://cloud-api.near.ai/v1",
    api_key=os.environ["NEARAI_API_KEY"],
)

reply = client.chat.completions.create(
    model="Qwen/Qwen3.8-27B-abliterated",
    messages=[{"role": "user", "content": "Explain how this SQL injection payload works."}],
)
print(reply.choices[0].message.content)
```

## How to choose an uncensored LLM API

- **Does it actually answer?** Test the prompts your work needs. NEAR AI Cloud: abliterated models, with the refusal direction removed from the weights.
- **Who can read your prompts?** On a typical API, "we don't log" is a promise. NEAR AI Cloud: the models run in a TEE, and you can verify it.
- **Does your code still work?** NEAR AI Cloud: OpenAI-compatible chat completions at cloud-api.near.ai/v1.
- **Can you start today?** NEAR AI Cloud: open to every account, no application, pay per token.
- **Are the rules clear?** NEAR AI Cloud: governed by a public Acceptable Use Policy.

## Security use cases

- **Vulnerability review** of your own code, without the model refusing to discuss the attack.
- **Authorized testing:** proof-of-concept exploits for systems you are authorized to test.
- **Threat analysis:** deobfuscate malware, explain payloads and read hostile bug bounty reports inside an enclave, then draft detection rules.
- **Red-teaming your AI:** generate adversarial prompts to test your own guardrails.

More: [Security research on NEAR AI Cloud](https://near.ai/cybersecurity). Security teams can join NEAR AI RED for discounts and early access.

## FAQ

### What is an uncensored LLM API?

It is an API that serves language models with their refusal behavior removed, so they answer requests a mainstream model would decline. NEAR AI Cloud serves abliterated models, where the refusal direction has been removed from the weights.

### Which uncensored models can I use through the API?

NEAR AI Cloud serves abliterated DeepSeek V4.1 Flash and Qwen 3.8 27B, each with its live context window and price on its model page.

### Is it OpenAI-compatible?

Yes. Point the OpenAI SDK, or any tool that accepts an OpenAI-compatible base URL, at https://cloud-api.near.ai/v1 and use your NEAR AI API key.

### Can NEAR AI read my prompts?

No. The abliterated models run inside trusted execution environments on NEAR AI's infrastructure, so NEAR AI and the cloud provider cannot see your prompts or the answers. You can verify it with a hardware attestation.

### Do I need to apply for access?

No. Every NEAR AI Cloud account can use the uncensored models, with no application and no identity checks. Security teams can join NEAR AI RED for discounts and early access.

### How much does the uncensored API cost?

You pay per token, with no minimums. The live price per million input and output tokens for each model is on each model page.

### Is there a free uncensored LLM API?

NEAR AI Cloud is pay as you go, so you pay only for the tokens you use. You can also download abliterated models from Hugging Face and run them on your own hardware for free, if you have a GPU that fits them.

### What am I allowed to use it for?

Legitimate work such as security research, authorized testing, malware analysis and red-teaming your own AI. Use is governed by NEAR AI's Acceptable Use Policy, and you stay responsible for what you do with the output.

### Can I use it with coding agents and frameworks?

Yes. Tools that let you set a custom OpenAI-compatible base URL, such as most coding agents, LangChain and LlamaIndex, work with NEAR AI Cloud.

By Alejo Escriva, reviewed by Sergey Astretsov. Updated November 20, 2026.

Canonical page: https://near.ai/uncensored-llm-api
