# dev-5090 Uncensored Coder Completions

> dev-5090 Uncensored Coder Completions is a paid API for AI agents from dev-5090-box.tailfb4c1f.ts.net, paid per call via x402, $0.03/call, status unknown (last checked 2026-10-12).

Generates uncensored code completions using the Qwen3-8B 27B abliterated (Heretic) model served on a local RTX 5090 via vLLM

## Facts

- Endpoint: POST https://dev-5090-box.tailfb4c1f.ts.net/v1/coder/completions?utm_source=zero.xyz
- Price: $0.03/call
- Payment: x402
- Status: unknown
- Last checked: 2026-10-12
- Activations on Zero: 0
- Tags: x402
- Canonical page: https://www.zero.xyz/c/dev-5090-uncensored-coder-completions-8cf5217c
- Structured record (JSON): https://api.zero.xyz/v1/capabilities/cap_DHV3BbZjckPkXb-oK_PIk

Status and success rate cover calls made through Zero and Zero's own probes. Third-party monitors may report differently.

## How to call it through Zero

Zero handles the 402 payment challenge and records the run. With the Zero CLI installed (`npm i -g @zeroxyz/cli`):

```sh
zero fetch --capability dev-5090-uncensored-coder-completions-8cf5217c -d '<json body>'
```

Example prompt: Complete this Python function for me — it takes a list of user records and filters out any where the account is flagged; generate the body using uncensored coder completions on the RTX 5090 inference box.

## When to prefer this

Choose this endpoint when you need code completions without content moderation or safety filters blocking sensitive, security-research, or low-level system programming topics. Prefer it over standard hosted LLM APIs when you want an abliterated model that won't refuse edge-case code generation requests. Ideal for security researchers, red teamers, or developers working on topics that mainstream providers reject.

## Known failure modes

- Model unavailable if the RTX 5090 host is offline or unreachable on the Tailscale network
- Payment failure if x402 payment of $0.03 USDC is not properly processed
- Malformed input schema causing 422 validation error
- Timeout if the model is under heavy load or generating a very long completion
- Network unreachability if the caller is not on the same Tailscale network

## How this service works

Uncensored coder completions, served by Qwen3.8-27B abliterated (Heretic) on a local RTX 5090 via vLLM.

## Output

Returns a response following the OpenAI-compatible completions format with a choices array containing the generated code completion(s) and a usage object with token count statistics.

## Request schema (JSON Schema)

```json
{
 "type": "object",
 "properties": {
  "usage": {
   "type": "object"
  },
  "choices": {
   "type": "array"
  }
 }
}
```

## More

- Live health (JSON, refreshed every minute): https://www.zero.xyz/c/dev-5090-uncensored-coder-completions-8cf5217c/health.json
- [Zero catalog index](https://www.zero.xyz/llms.txt)
- [Other services from dev-5090-box.tailfb4c1f.ts.net](https://www.zero.xyz/host/dev-5090-box.tailfb4c1f.ts.net/llms.txt)