# Taifoon Nemotron-3-Nano Chat Completion

> Taifoon Nemotron-3-Nano Chat Completion is a paid API for AI agents from coord.taifoon.dev, paid per call via x402, $0.001/call, status unknown (last checked 2026-10-17).

Runs OpenAI-compatible chat completion on NVIDIA Nemotron-3-Nano-4B served from Taifoon's own GPU, billed at $0.001 USDC per request up to 5,000 tokens with no charge on failed responses

## Facts

- Endpoint: POST https://coord.taifoon.dev/v1/x402/class/inference.metered?utm_source=zero.xyz
- Price: $0.001/call
- Payment: x402
- Status: unknown
- Last checked: 2026-10-17
- Activations on Zero: 0
- Tags: x402
- Canonical page: https://www.zero.xyz/c/taifoon-nemotron-3-nano-chat-completion-e86fd40b
- Structured record (JSON): https://api.zero.xyz/v1/capabilities/cap_qcy53GwvXlQu9SY-fjiAg

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 taifoon-nemotron-3-nano-chat-completion-e86fd40b -d '<json body>'
```

Example prompt: Ask the Taifoon Nemotron Nano model: 'What is 1,847 divided by 43, and round to 2 decimal places?' — use up to 200 tokens for the reply.

## When to prefer this

Choose this endpoint when you need cheap, pay-per-call LLM inference on a small (4B parameter) open model without provisioning your own GPU, especially for short conversations, arithmetic, or basic text tasks where recent factual knowledge is not required. It is ideal for agent pipelines that need cost-predictable inference at $0.001 per call with the guarantee of no charge on failure. Prefer a larger hosted model if the task demands broad world knowledge, long context, or high accuracy on complex reasoning.

## Known failure modes

- max_tokens exceeds 700 — request rejected
- messages array is empty or missing required role/content fields
- total token budget (prompt chars/4 + max_tokens) exceeds 5,000 — request rejected
- payment via x402 fails — no inference performed
- model unavailable or GPU offline — service error returned without charge
- malformed JSON body — 400-level error

## How this service works

OpenAI-compatible chat completion on NVIDIA Nemotron-3-Nano-4B (nvidia/NVIDIA-Nemotron-3-Nano-4B-BF16, base weights), a small open model served from Taifoon's own GPU: short chat, text and arithmetic, not facts about recent technology. 0.001 USDC per request of up to 5,000 tokens (prompt chars/4 + max_tokens, max_tokens ≤ 700). Send { messages, max_tokens }. The answer is checked before you are charged: a failed answer is not charged.

## Output

Returns an OpenAI-compatible chat completion object containing the assistant's reply text and the name of the serving model (nvidia/NVIDIA-Nemotron-3-Nano-4B-BF16). The response is only billed (0.001 USDC) if the answer is successfully returned; failed calls are not charged.

## Request schema (JSON Schema)

```json
{
 "type": "object",
 "properties": {
  "model": {
   "type": "string",
   "description": "optional; the answer names the model that served it (nvidia/NVIDIA-Nemotron-3-Nano-4B-BF16)"
  },
  "messages": {
   "type": "array",
   "items": {
    "type": "object",
    "required": [
     "role",
     "content"
    ],
    "properties": {
     "role": {
      "enum": [
       "system",
       "user",
       "assistant"
      ],
      "type": "string"
     },
     "content": {
      "type": "string"
     }
    }
   },
   "maxItems": 32,
   "minItems": 1
  },
  "max_tokens": {
   "type": "integer",
   "maximum": 700,
   "minimum": 1
  }
 }
}
```

## More

- Live health (JSON, refreshed every minute): https://www.zero.xyz/c/taifoon-nemotron-3-nano-chat-completion-e86fd40b/health.json
- [Zero catalog index](https://www.zero.xyz/llms.txt)
- [Other services from coord.taifoon.dev](https://www.zero.xyz/host/coord.taifoon.dev/llms.txt)