Hormouz AIHormouzSign in

NVIDIA: Nemotron 3 Ultra: AI model

Nemotron 3 Ultra is NVIDIA's open reasoning-and-orchestration model with 55B active parameters (of 550B total) and a context window up to 1M tokens. Reachable from inside Iran through Hormouz.

Context window:
262,144 tokens
Input:
$0.6/1M tokens
Output:
$2.4/1M tokens
Type:
text
  • Tools
  • Streaming
  • Long context

Thinking levels: high · medium

Nemotron 3 Ultra is an open frontier reasoning-and-orchestration model from NVIDIA, built on a hybrid Transformer-Mamba Mixture-of-Experts architecture that activates 55B of its 550B parameters. It takes text input and output and supports a context window of up to 1M tokens.

It suits long-running agentic workflows. Agent orchestration, coding agents, deep research and complex enterprise tasks. And is particularly strong at multi-step reasoning and planning, with high-throughput inference designed for high-volume agent pipelines. Hormouz exposes an OpenAI-compatible API. Point your base URL at Hormouz, drop in a Hormouz key, and existing OpenAI SDK code runs unchanged.

Pricing is usage-based: $1/M in · $3.6/M out. Hormouz provides stable access to the model from inside Iran, with no foreign credit card or circumvention tooling required.

NVIDIA Nemotron 3 Ultra is an open frontier-reasoning and orchestration model from NVIDIA, with 55B active parameters out of 550B total (MoE). Built on a hybrid Transformer-Mamba mixture-of-experts architecture, it...

Quickstart

curl https://api.hormouz.net/v1/chat/completions \
  -H "Authorization: Bearer ng-YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "nvidia/nemotron-3-ultra-550b-a55b",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

Specs

VendorNVIDIA
TypeText
Context window262,144 tokens
Pricing$0.60/M in · $2.40/M out
APIChat Completions
CapabilitiesStreaming · Tools

Benchmarks

22.9
Intelligence Index
Better than 41% of models we compared
49.3
Coding Index
Better than 53% of models we compared
20.1
Agentic Index
Better than 49% of models we compared

Reasoning

GPQA Diamond
Graduate-level scientific reasoning
86.7%
HLE
Humanity's Last Exam
28.4%
IFBench
Instruction-following
81.4%
AA-LCR
Long-context reasoning
79.3%
CritPt
Research-level physics reasoning
3.1%
AA Omniscience
Broad factual knowledge accuracy
22.6%
Non-Hallucination
Rate of avoiding fabricated answers
70.3%

Coding

SciCode
Python for scientific computing
40.3%
Terminal-Bench Hard
Agentic coding & terminal use
36.4%

Agentic

τ²-Bench
Tool-use agents in dual-control scenarios
83.3%
GDPval
Economically valuable tasks
25.8%

Source: Artificial Analysis

Design Arena

How Design Arena works ↗
3D
Top 53%
1,141Elo
Win rate: 39.2%Avg time: 106s
ASCII Art
Top 85%
1,105Elo
Win rate: 38%Avg time: 103s
Code
Top 63%
1,150Elo
Win rate: 36.2%Avg time: 132s
Data Viz
Top 63%
1,150Elo
Win rate: 37.7%Avg time: 108s
Game Dev
Top 56%
1,152Elo
Win rate: 36.8%Avg time: 122s
SVG
Top 76%
1,069Elo
Win rate: 32.8%Avg time: 39.2s
UI Component
Top 63%
1,146Elo
Win rate: 36.4%Avg time: 101s
Website
Top 66%
1,146Elo
Win rate: 34.9%Avg time: 149s

Metrics sourced from Design Arena

Code example

cURL
curl https://api.hormouz.net/v1/chat/completions \
  -H "Authorization: Bearer ng-YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "nvidia/nemotron-3-ultra-550b-a55b",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'
Python
from openai import OpenAI
client = OpenAI(api_key="ng-YOUR_API_KEY", base_url="https://api.hormouz.net/v1")
response = client.chat.completions.create(
    model="nvidia/nemotron-3-ultra-550b-a55b",
    messages=[{"role": "user", "content": "Hello!"}],
)
print(response.choices[0].message.content)
JavaScript
import OpenAI from "openai";
const client = new OpenAI({ apiKey: "ng-YOUR_API_KEY", baseURL: "https://api.hormouz.net/v1" });
const response = await client.chat.completions.create({
  model: "nvidia/nemotron-3-ultra-550b-a55b",
  messages: [{ role: "user", content: "Hello!" }],
});
console.log(response.choices[0].message.content);

Buy the NVIDIA: Nemotron 3 Ultra API, paid in rial

  • Buy credit with an Iranian bank card; no foreign card or account needed.
  • One Hormouz key reaches NVIDIA: Nemotron 3 Ultra and 600+ other models, and you pay only for what you use.
  • No VPN and no risk of a banned account; just change the base URL to api.hormouz.net/v1.
Create an account

Use via Hormouz

Connect to this model and 600+ others with one API key, then see usage down to each request.

Request company access

FAQ

How much does Nemotron 3 Ultra cost?
Billing is usage-based: $1/M in · $3.6/M out. There is no fixed monthly fee. You pay only for what you use.
How do I use Nemotron 3 Ultra from inside Iran?
Create a Hormouz account and generate an API key. Hormouz gives you a stable route to the model from inside Iran, with no foreign credit card or circumvention tooling needed.
Is Nemotron 3 Ultra OpenAI-compatible?
Yes. Hormouz serves an OpenAI-compatible API. Point your base URL at Hormouz and swap in a Hormouz key; your existing SDKs and code work without a rewrite.
How large is Nemotron 3 Ultra's context window?
It supports a context window of up to 1M tokens, suiting very long documents and long-horizon agentic work.
What is Nemotron 3 Ultra good for?
Agent orchestration, coding agents, deep research and multi-step enterprise tasks that need strong reasoning and planning.
How does this differ from the free tier?
A free, rate-limited version of the same model exists; for production and high volume this paid tier is the more stable choice.

Updated 2026-10-13