# Qwen3-27B Multi-Step Reasoning & Planning API

> Qwen3-27B Multi-Step Reasoning & Planning API is a paid API for AI agents from linked-aimed-reproduced-changing.trycloudflare.com, paid per call via x402, $0.05/call, status unknown (last checked 2026-10-12).

Runs multi-step reasoning and planning tasks using an uncensored Qwen3.8-27B model served locally via vLLM on an RTX 5090

## Facts

- Endpoint: POST https://linked-aimed-reproduced-changing.trycloudflare.com/v1/reason?utm_source=zero.xyz
- Price: $0.05/call
- Payment: x402
- Status: unknown
- Last checked: 2026-10-12
- Activations on Zero: 0
- Tags: x402
- Canonical page: https://www.zero.xyz/c/qwen3-27b-multi-step-reasoning-planning-api-e7a735d4
- Structured record (JSON): https://api.zero.xyz/v1/capabilities/cap_757cqdC-_4QgltN8qJ-lr

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 qwen3-27b-multi-step-reasoning-planning-api-e7a735d4 -d '<json body>'
```

Example prompt: Use the multi-step reasoning endpoint to work through this problem: given three interdependent software modules with circular dependencies, plan a refactoring strategy that breaks the cycles while preserving all public APIs.

## When to prefer this

Choose this endpoint when you need genuine multi-step reasoning or chain-of-thought planning, especially on problems that benefit from an uncensored model (edge cases, security research, adversarial scenarios). Prefer it over standard chat completions when the task requires decomposition into sub-steps rather than a single-shot answer. The local RTX 5090 hardware means low-latency inference for moderately long reasoning chains without cloud API rate limits.

## Known failure modes

- Payment not received or x402 protocol failure — request rejected before inference
- Model overloaded on local GPU — may return 503 or timeout under heavy concurrent load
- Prompt too long for context window — returns error or truncated response
- Cloudflare Tunnel instability causing connection drops mid-stream
- Abliterated model may produce lower-quality refusals for truly malformed inputs rather than structured errors

## How this service works

Multi-step reasoning and planning, served by Qwen3.8-27B abliterated (Heretic) on a local RTX 5090 via vLLM.

## Output

Returns an OpenAI-compatible response object containing a reasoning chain or structured plan as the assistant message content, along with usage statistics and choices array reflecting the model's multi-step thought process.

## 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/qwen3-27b-multi-step-reasoning-planning-api-e7a735d4/health.json
- [Zero catalog index](https://www.zero.xyz/llms.txt)
- [Other services from linked-aimed-reproduced-changing.trycloudflare.com](https://www.zero.xyz/host/linked-aimed-reproduced-changing.trycloudflare.com/llms.txt)