# Nemotron 3 120B (wafermap.xyz)

> Nemotron 3 120B (wafermap.xyz) is a paid API for AI agents from wafermap.xyz, paid per call via x402, $0.01/call, status unknown (last checked 2026-10-31).

Runs NVIDIA's Nemotron 3 120B mixture-of-experts language model, tuned for agentic workflows, on a pay-per-call basis via x402.

## Facts

- Endpoint: POST https://wafermap.xyz/v1/nemotron-3?utm_source=zero.xyz
- Price: $0.01/call
- Payment: x402
- Status: unknown
- Last checked: 2026-10-31
- Activations on Zero: 0
- Tags: x402
- Canonical page: https://www.zero.xyz/c/nemotron-3-120b-wafermap-xyz-8832ab4d
- Structured record (JSON): https://api.zero.xyz/v1/capabilities/cap_BXQvEjROejphArc_--9Yo

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 nemotron-3-120b-wafermap-xyz-8832ab4d -d '<json body>'
```

Example prompt: Ask Nemotron 3 120B to help me plan a multi-step research workflow: use the system prompt 'You are a rigorous research assistant who breaks tasks into clear steps' and the user prompt 'I need to compare three open-source vector databases for a production RAG system — outline evaluation criteria, trade-offs, and a recommended decision framework', with a 1500 token output cap.

## When to prefer this

Choose this endpoint when you need a large, open MoE model specifically tuned for agentic and instruction-following workflows, particularly if you want NVIDIA's Nemotron 3 120B rather than alternatives like DeepSeek or Llama. It is well-suited for multi-step reasoning, tool-use planning, and structured output generation. The pay-per-call x402 model makes it convenient for agent pipelines without subscription overhead. Prefer DeepSeek V4 Pro for harder math/code, or Llama 4 Scout for multimodal tasks.

## Known failure modes

- Prompt exceeds 24000 character limit — request rejected
- max_tokens set to 0 or negative — validation error
- System prompt exceeds 8000 character limit — request rejected
- Payment not provided or insufficient USDC — x402 payment required error
- Model unavailable or overloaded — 503 or timeout response
- Output truncated if max_tokens is set too low for the task

## How this service works

Nemotron 3 120B — NVIDIA's open MoE model tuned for agent workflows.

## Output

Returns a generated text completion from the Nemotron 3 120B MoE model, up to 2048 tokens. The response reflects the provided system prompt instructions and answers or expands on the user's prompt, optimized for agentic and instruction-following use cases.

## Request schema (JSON Schema)

```json
{
 "type": "object",
 "properties": {
  "prompt": {
   "type": "string",
   "maxLength": 24000,
   "minLength": 1,
   "description": "What you want, in plain words."
  },
  "system": {
   "type": "string",
   "maxLength": 8000,
   "minLength": 1,
   "description": "Optional instructions that shape every answer."
  },
  "max_tokens": {
   "type": "number",
   "maximum": 2048,
   "description": "Output cap. 2048 at most, 1024 by default.",
   "exclusiveMinimum": 0
  }
 }
}
```

## More

- Live health (JSON, refreshed every minute): https://www.zero.xyz/c/nemotron-3-120b-wafermap-xyz-8832ab4d/health.json
- [Zero catalog index](https://www.zero.xyz/llms.txt)
- [Other services from wafermap.xyz](https://www.zero.xyz/host/wafermap.xyz/llms.txt)