Nemotron 3 Ultra

by nvidia · family nemotron

Facts

Release date	2026-06-04
Retired date	—
Knowledge cutoff	—
Context (max)	1,000,000
Max output	—
Modalities	text → text
Aliases	`nvidia/nemotron-3-ultra-550b-a55bNVIDIA-Nemotron-3-Ultra-550B-A55B`
Open weights	yes
Description	550B MoE (55B active) reasoning model; served on NVIDIA NIM and Baseten.

Unknown values are shown as —.

Intelligence Index v4.1.1: 38 · 550B-A55B · Source: Artificial Analysis (accessed 2026-08-20)

Offerings

Provider	Wire ID	Endpoint	Protocol	Status	Price (USD / Mtok)	Reasoning
NVIDIA NIM API	`nvidia/nemotron-3-ultra-550b-a55b`	v1-chat-completions	openai-chat	ga	—	toggle · default on · returns reasoning_content
Ollama Cloud	`nemotron-3-ultra`	v1-chat-completions	openai-chat	ga	—	effort · default on · returns hidden
OpenCode Zen	`nemotron-3-ultra-free`	chat-completions	openai-chat	ga	—	toggle · default on · returns reasoning_content
OpenRouter	`nvidia/nemotron-3-ultra-550b-a55b`	chat-completions	openai-chat	ga	0.60 / 3.60	effort · default on · returns reasoning_content · round-trips reasoning_content

Reasoning controls

NVIDIA NIM API· nvidia/nemotron-3-ultra-550b-a55b· v1-chat-completions toggle

chat_template_kwargs.enable_thinking on true · off false

toggle · default on · returns reasoning_content

Ollama Cloud· nemotron-3-ultra· v1-chat-completions effort

reasoning_effort nonelowmediumhigh●max Default level not documented; verify.

effort · default on · returns hidden

OpenCode Zen· nemotron-3-ultra-free· chat-completions toggle

chat_template_kwargs.enable_thinking on true · off false

toggle · default on · returns reasoning_content

OpenRouter· nvidia/nemotron-3-ultra-550b-a55b· chat-completions effort

reasoning.effort nonelowmedium●highxhighmax

effort · default on · returns reasoning_content · round-trips reasoning_content

Example requests

NVIDIA NIM API · openai-chat · nvidia/nemotron-3-ultra-550b-a55b

curl 'https://integrate.api.nvidia.com/v1/chat/completions' \
  -H 'Authorization: Bearer <YOUR_API_KEY>' \
  -H 'Content-Type: application/json' \
  -d '{
  "model": "nvidia/nemotron-3-ultra-550b-a55b",
  "messages": [
    {
      "role": "user",
      "content": "Tell me about the weather."
    }
  ],
  "chat_template_kwargs": {
    "enable_thinking": true
  }
}'

Python (requests)

import requests

resp = requests.post(
    "https://integrate.api.nvidia.com/v1/chat/completions",
    headers={
      "Authorization": "Bearer <YOUR_API_KEY>",
      "Content-Type": "application/json"
    },
    json={
      "model": "nvidia/nemotron-3-ultra-550b-a55b",
      "messages": [
        {
          "role": "user",
          "content": "Tell me about the weather."
        }
      ],
      "chat_template_kwargs": {
        "enable_thinking": true
      }
    },
)
print(resp.json())

TypeScript (fetch)

const resp = await fetch("https://integrate.api.nvidia.com/v1/chat/completions", {
  method: "POST",
  headers: {
    "Authorization": "Bearer <YOUR_API_KEY>",
    "Content-Type": "application/json"
  },
  body: JSON.stringify({
    "model": "nvidia/nemotron-3-ultra-550b-a55b",
    "messages": [
      {
        "role": "user",
        "content": "Tell me about the weather."
      }
    ],
    "chat_template_kwargs": {
      "enable_thinking": true
    }
  }),
})
console.log(await resp.json())

Ollama Cloud · openai-chat · nemotron-3-ultra

curl 'https://ollama.com/v1/chat/completions' \
  -H 'Authorization: Bearer <YOUR_API_KEY>' \
  -H 'Content-Type: application/json' \
  -d '{
  "model": "nemotron-3-ultra",
  "messages": [
    {
      "role": "user",
      "content": "Tell me about the weather."
    }
  ],
  "reasoning_effort": "high"
}'

Python (requests)

import requests

resp = requests.post(
    "https://ollama.com/v1/chat/completions",
    headers={
      "Authorization": "Bearer <YOUR_API_KEY>",
      "Content-Type": "application/json"
    },
    json={
      "model": "nemotron-3-ultra",
      "messages": [
        {
          "role": "user",
          "content": "Tell me about the weather."
        }
      ],
      "reasoning_effort": "high"
    },
)
print(resp.json())

TypeScript (fetch)

const resp = await fetch("https://ollama.com/v1/chat/completions", {
  method: "POST",
  headers: {
    "Authorization": "Bearer <YOUR_API_KEY>",
    "Content-Type": "application/json"
  },
  body: JSON.stringify({
    "model": "nemotron-3-ultra",
    "messages": [
      {
        "role": "user",
        "content": "Tell me about the weather."
      }
    ],
    "reasoning_effort": "high"
  }),
})
console.log(await resp.json())

OpenCode Zen · openai-chat · nemotron-3-ultra-free

curl 'https://opencode.ai/zen/v1/chat/completions' \
  -H 'Authorization: Bearer <YOUR_API_KEY>' \
  -H 'Content-Type: application/json' \
  -d '{
  "model": "nemotron-3-ultra-free",
  "messages": [
    {
      "role": "user",
      "content": "Tell me about the weather."
    }
  ],
  "chat_template_kwargs": {
    "enable_thinking": true
  }
}'

Python (requests)

import requests

resp = requests.post(
    "https://opencode.ai/zen/v1/chat/completions",
    headers={
      "Authorization": "Bearer <YOUR_API_KEY>",
      "Content-Type": "application/json"
    },
    json={
      "model": "nemotron-3-ultra-free",
      "messages": [
        {
          "role": "user",
          "content": "Tell me about the weather."
        }
      ],
      "chat_template_kwargs": {
        "enable_thinking": true
      }
    },
)
print(resp.json())

TypeScript (fetch)

const resp = await fetch("https://opencode.ai/zen/v1/chat/completions", {
  method: "POST",
  headers: {
    "Authorization": "Bearer <YOUR_API_KEY>",
    "Content-Type": "application/json"
  },
  body: JSON.stringify({
    "model": "nemotron-3-ultra-free",
    "messages": [
      {
        "role": "user",
        "content": "Tell me about the weather."
      }
    ],
    "chat_template_kwargs": {
      "enable_thinking": true
    }
  }),
})
console.log(await resp.json())

OpenRouter · openai-chat · nvidia/nemotron-3-ultra-550b-a55b

curl 'https://openrouter.ai/api/v1/chat/completions' \
  -H 'Authorization: Bearer <YOUR_API_KEY>' \
  -H 'Content-Type: application/json' \
  -d '{
  "model": "nvidia/nemotron-3-ultra-550b-a55b",
  "messages": [
    {
      "role": "user",
      "content": "Tell me about the weather."
    }
  ],
  "reasoning": {
    "effort": "medium"
  }
}'

Python (requests)

import requests

resp = requests.post(
    "https://openrouter.ai/api/v1/chat/completions",
    headers={
      "Authorization": "Bearer <YOUR_API_KEY>",
      "Content-Type": "application/json"
    },
    json={
      "model": "nvidia/nemotron-3-ultra-550b-a55b",
      "messages": [
        {
          "role": "user",
          "content": "Tell me about the weather."
        }
      ],
      "reasoning": {
        "effort": "medium"
      }
    },
)
print(resp.json())

TypeScript (fetch)

const resp = await fetch("https://openrouter.ai/api/v1/chat/completions", {
  method: "POST",
  headers: {
    "Authorization": "Bearer <YOUR_API_KEY>",
    "Content-Type": "application/json"
  },
  body: JSON.stringify({
    "model": "nvidia/nemotron-3-ultra-550b-a55b",
    "messages": [
      {
        "role": "user",
        "content": "Tell me about the weather."
      }
    ],
    "reasoning": {
      "effort": "medium"
    }
  }),
})
console.log(await resp.json())

⚠ Responses include reasoning_content artifacts — pass them back unmodified on subsequent turns (dropping them triggers 400s or quality loss).