# Flat-rate LLM Inference

> Flat-rate LLM Inference is a paid API for AI agents from flat-rate-llm.kikoribera03.workers.dev, paid per call via x402, $1/call, status unknown (last checked 2026-09-15).

Runs LLM inference on a prompt at a fixed $0.001 per call regardless of token count, with automatic failover across multiple large models.

## Facts

- Endpoint: POST https://flat-rate-llm.kikoribera03.workers.dev/v2/inferencia
- Price: $1/call
- Payment: x402
- Status: unknown
- Last checked: 2026-09-15
- Activations on Zero: 0
- Tags: x402
- Canonical page: https://www.zero.xyz/c/flat-rate-llm-inference-479a3149
- Structured record (JSON): https://api.zero.xyz/v1/capabilities/cap_4DsbNO-kJg0kr_PdumrSN

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 flat-rate-llm-inference-479a3149 -d '<json body>'
```

Example prompt: Send this prompt to the flat-rate LLM with the system instruction 'You are a helpful assistant' and cap the output at 500 tokens: 'Summarize the key differences between REST and GraphQL APIs in plain English.'

## When to prefer this

Choose this endpoint when you need predictable, fixed-cost LLM inference regardless of prompt or output length — ideal for agentic pipelines making many calls, batch workflows where token-metered APIs create variable cost, or cost-sensitive automation. Prefer it over token-metered gateways when throughput is high or prompt lengths vary widely. Not ideal if you need specific model selection, streaming output, or detailed token usage telemetry.

## Known failure modes

- Prompt exceeds 24000 character limit — request rejected
- maxTokens set above 4000 — capped or rejected
- Model failover exhausted — service temporarily unavailable with 5xx error
- Payment not included or invalid x402 payment — 402 Payment Required
- Malformed JSON body — 400 Bad Request

## How this service works

60 paid endpoints with no metering, no API key and no account. Each endpoint has one flat price per call, whatever the size of the request: LLM completions with automatic failover across several large models, read-only EVM tooling over Base, Ethereum, Polygon, Arbitrum and Optimism, and pure crypto utilities that touch no network.

## Output

A text string containing the model's completion of the given prompt, generated by one of the backend large language models selected via automatic failover. No token usage metadata is returned; pricing is flat per call.

## Request schema (JSON Schema)

```json
{
 "type": "object",
 "required": [
  "prompt"
 ],
 "properties": {
  "prompt": {
   "type": "string",
   "description": "The request for the model. Up to 24000 characters."
  },
  "sistema": {
   "type": "string",
   "description": "Optional system instruction."
  },
  "maxTokens": {
   "type": "number",
   "description": "Output token cap, up to 4000. Does not change the price."
  }
 }
}
```

## Response schema (JSON Schema)

```json
{
 "type": "object",
 "properties": {
  "modelo": {
   "type": "string"
  },
  "segundos": {
   "type": "number"
  },
  "respuesta": {
   "type": "string"
  }
 }
}
```

## More

- Live health (JSON, refreshed every minute): https://www.zero.xyz/c/flat-rate-llm-inference-479a3149/health.json
- [Zero catalog index](https://www.zero.xyz/llms.txt)
- [Other services from flat-rate-llm.kikoribera03.workers.dev](https://www.zero.xyz/host/flat-rate-llm.kikoribera03.workers.dev/llms.txt)