impulse

Nemotron 3 Ultra 550B A55B

Open weights

NVIDIA Nemotron 3 Ultra is an open frontier-reasoning and orchestration model from NVIDIA, with 55B active parameters out of 550B total (MoE).

API id nemotron-3-ultra-550b-a55bReleased 4 Jun 2026 (Jun 2026)
Input / 1M
-
not listed
Output / 1M
-
not listed
Context window
1M
tokens
Max output
33K
tokens per reply
Hosts
5
5 listings
Released
Jun 2026
4 Jun 2026

Where to run it

5 hosts serve Nemotron 3 Ultra 550B A55B. The comparison is on blended price, 3 parts input to 1 part output.

HostIn / out per 1M
NVIDIA NIMInference
nemotron-3-ultra-550b-a55b
-
PerplexityGateway
nemotron-3-ultra-550b-a55b
$0.25 / $2.50
GMI CloudInference
nemotron-3-ultra-550b-a55b
$0.80 / $2.60
Together AIInference
nemotron-3-ultra-550b-a55b
$0.60 / $3.60
Nemotron-3-Ultra-550b-a55b
$1.00 / $3.00

Capabilities

Core

  • Streaming
  • Function calling
  • JSON mode
  • Structured outputs

Multimodal

  • Vision
  • Image input
  • Audio
  • Image generation
  • Video

Advanced

  • Reasoning
  • Tool calling
  • MCP
  • Prompt caching
  • Batch API
  • Embeddings
Takes in
Text
Produces
Text
Open weights
License: NVIDIA Open Model License
Runs in about 1345.3 GB of GPU memory
Runtimes: TGI, vLLM

Similar models

This model has no listed price, so these are current models with similar capabilities.

Smaller NVIDIA models

No similar NVIDIA model keeps its key capabilities.

Other frontier labs

  • GPT-5.5 Pro
    OpenAI · on OpenAI
    $30.00 / $180.00
    Has every key capability
  • Claude Fable 5.2
    Anthropic · on Anthropic
    $8.00 / $40.00
    Has every key capability
  • Gemini 4 Carbon
    Google · on Google AI Studio
    $2.00 / $10.00
    Has every key capability
  • Grok 4.7
    xAI · on xAI
    $2.00 / $6.00
    Has every key capability

Open-weight picks

  • Command A Plus
    Cohere · on AIHubMix
    $2.50 / $10.00
    Has every key capability
  • Kimi K3.1 Pro
    Moonshot AI · on Databricks
    $2.00 / $10.00
    Has every key capability
  • Qwen3.8.2 4T A95B
    Alibaba Qwen · on DeepInfra
    $2.00 / $6.00
    Has every key capability
  • Muse Spark Open
    Meta · on Baseten
    $1.25 / $4.25
    Has every key capability

Call it

Using NVIDIA NIM's API.

Python
from litellm import completion

response = completion(
    model="nvidia_nim/nemotron-3-ultra-550b-a55b",
    messages=[{"role": "user", "content": "Hello"}],
)
print(response.choices[0].message.content)

Quick answers

How much does Nemotron 3 Ultra 550B A55B cost?

Its lowest listed price is at Perplexity: $0.25 per million input tokens and $2.50 per million output tokens.

Which APIs offer Nemotron 3 Ultra 550B A55B?

Nemotron 3 Ultra 550B A55B runs on 5 APIs: NVIDIA NIM, Perplexity, GMI Cloud, Together AI and Nebius AI Studio.

What is Nemotron 3 Ultra 550B A55B's context window?

1M tokens, with up to 33K tokens in one reply.

Is Nemotron 3 Ultra 550B A55B being retired?

No retirement date is scheduled for it on any API we track.

When was Nemotron 3 Ultra 550B A55B released?

NVIDIA released it on 4 Jun 2026.

What is Nemotron 3 Ultra 550B A55B's API model id?

Use nemotron-3-ultra-550b-a55b with NVIDIA NIM. Other hosts use their own ids: Nemotron-3-Ultra-550b-a55b on Nebius AI Studio.

Sources

Raw catalog entry for nemotron-3-ultra-550b-a55b on NVIDIA NIM
{
  "litellm_provider": "nvidia_nim",
  "max_output_tokens": 32768,
  "mode": "chat",
  "source": "https://build.nvidia.com/nvidia/nemotron-3-ultra-550b-a55b.md",
  "supported_endpoints": [
    "/v1/chat/completions"
  ]
}