# SableForge LLM — Locally Hosted 27B Language Model via x402

> SableForge LLM — Locally Hosted 27B Language Model via x402 is a paid API for AI agents from sableforge.servehttp.com, paid per call via x402, $0.002/call, status unknown (last checked 2026-10-11).

Runs a locally hosted Qwen3 27B large language model inference endpoint, billed per-call in USDC on Base via x402 v2 with no API keys or accounts required.

## Facts

- Endpoint: POST https://sableforge.servehttp.com/api/llm?utm_source=zero.xyz
- Price: $0.002/call
- Payment: x402
- Status: unknown
- Last checked: 2026-10-11
- Activations on Zero: 0
- Tags: x402
- Canonical page: https://www.zero.xyz/c/sableforge-llm-locally-hosted-27b-language-model-via-x402-5c0654cf
- Structured record (JSON): https://api.zero.xyz/v1/capabilities/cap_M4k4fDmaReiUeoPYNHk6E

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 sableforge-llm-locally-hosted-27b-language-model-via-x402-5c0654cf -d '<json body>'
```

Example prompt: Send this conversation to SableForge's locally hosted 27B Qwen model and get a response: system message is 'You are a helpful assistant', user message is 'Explain the concept of zero-knowledge proofs in plain English.'

## When to prefer this

Choose this endpoint when you need pay-per-call LLM inference with no API key registration, no monthly subscription, and anonymous access settled in USDC on Base. It is especially well-suited for AI agents that hold crypto wallets and need to invoke LLM reasoning on demand. Prefer it over OpenAI or Anthropic APIs when avoiding account creation is a hard requirement, or when operating in a fully on-chain / x402 payment context. The 27B parameter size (Qwen3) provides strong general reasoning at a fraction of frontier model cost.

## Known failure modes

- Payment not settled — 402 response returned if USDC payment via x402 v2 is not included or fails
- Malformed messages array — missing role or content fields cause a 400 error
- Model overload — locally hosted hardware may introduce latency spikes or 503 under high concurrency
- Empty or null content — if messages array is empty, response may be undefined or error
- Network unavailability — servehttp.com dynamic DNS may have uptime variability

## How this service works

12 pay-per-call HTTP APIs settled in USDC on Base via x402 v2. No API keys, no accounts, no signup — $0.0005–$0.0020 per call. Includes a locally hosted 27B LLM endpoint.

## Output

A JSON object containing the model identifier (e.g. 'qwen3.8:27b') and the generated text content string from the LLM, returned synchronously after inference completes.

## Request schema (JSON Schema)

```json
{
 "type": "object",
 "properties": {
  "messages": {
   "type": "array",
   "items": {
    "type": "object",
    "properties": {
     "role": {
      "type": "string"
     },
     "content": {
      "type": "string"
     }
    }
   }
  }
 }
}
```

## Response schema (JSON Schema)

```json
{
 "type": "json",
 "example": {
  "model": "qwen3.8:27b",
  "content": "Hello!"
 }
}
```

## More

- Live health (JSON, refreshed every minute): https://www.zero.xyz/c/sableforge-llm-locally-hosted-27b-language-model-via-x402-5c0654cf/health.json
- [Zero catalog index](https://www.zero.xyz/llms.txt)
- [Other services from sableforge.servehttp.com](https://www.zero.xyz/host/sableforge.servehttp.com/llms.txt)