Requesty
Fireworks AI

nemotron-3-ultra-nvfp4

Nemotron-3-Ultra-550B-A55B-NVFP4 is a frontier-scale large language model (LLM) trained by NVIDIA, designed to deliver strong agentic, reasoning, and conversational capabilities. It is optimized for the most demanding workloads, including complex multi-step agents, long-context analysis, and high-accuracy reasoning over code, math, and science. The model employs a hybrid Latent Mixture-of-Experts (LatentMoE) architecture, utilizing interleaved Mamba-2 and MoE layers, along with select Attention layers. Like the Super model, the Ultra model incorporates Multi-Token Prediction (MTP) layers for faster text generation and improved quality, and it is trained using an NVFP4 pre-training recipe to maximize compute efficiency. The model has 55B active parameters and 550B parameters in total.

ReasoningTool callingCaching
Try nemotron-3-ultra-nvfp4CompareDocs

Specifications

Context window262K tokens
Max output262K tokens
API typechat
AddedJun 13, 2026
Model IDfireworks/nemotron-3-ultra-nvfp4
Data retentionNo
Used for trainingNo
Provider location🇺🇸 US
Privacy policyFireworks Privacy Policy→

Benchmarks

Benchmarks haven't been published yet for this exact variant.

Some variants (region-specific deployments, highspeed tiers) share benchmarks with their base model. Check the base model page or the Fireworks AI models overview.

Pricing

Input / 1M
$0.60
Output / 1M
$2.40
Cache write
N/A
Cache read / 1M
$0.12
Estimated cost
100K input + 10K output$0.0840
1M input + 100K output$0.84
10M input + 1M output$8.40

Requesty charges exactly what the upstream provider charges, no markup, no per-request fees. Prompt caching and smart routing can reduce effective cost by 30-80%.

Quickstart

Drop-in compatible with the OpenAI SDK. Change the base URL, swap in your Requesty API key, and set the model to fireworks/nemotron-3-ultra-nvfp4.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
from openai import OpenAI

client = OpenAI(
    api_key="YOUR_REQUESTY_API_KEY",
    base_url="https://router.requesty.ai/v1",
)

response = client.chat.completions.create(
    model="fireworks/nemotron-3-ultra-nvfp4",
    messages=[
        {"role": "user", "content": "Explain quantum computing in one paragraph."},
    ],
)

print(response.choices[0].message.content)

Other Fireworks AI models

qwen3.7-plus
262K ctx$0.40/1M
Coding 47
minimax-m3
512K ctx$0.30/1M
Coding 43
kimi-k2.7-code
262K ctx$0.95/1M
gpt-oss-120b
131K ctx$0.15/1M
Coding 29
gpt-oss-20b
131K ctx$0.07/1M
Coding 19
deepseek-v4-flash
1M ctx$0.14/1M
Coding 39

Frequently asked questions

How much does nemotron-3-ultra-nvfp4 cost?
nemotron-3-ultra-nvfp4 is priced at $0.60 per million input tokens and $2.40 per million output tokens when accessed via Requesty. Prompt caching is supported, which can cut effective input cost by up to 90% on repeated context. Requesty charges exactly what the upstream provider charges, we don't add markup.
What is the context window of nemotron-3-ultra-nvfp4?
nemotron-3-ultra-nvfp4 has a context window of 262K tokens, with a maximum output of 262K tokens per response. That's roughly 350 words of input you can fit in a single prompt.
What can nemotron-3-ultra-nvfp4 do?
nemotron-3-ultra-nvfp4 supports tool calling, extended reasoning, prompt caching. You can call it through any OpenAI-compatible client by pointing base_url to Requesty.
How do I use nemotron-3-ultra-nvfp4 with the OpenAI SDK?
Install the OpenAI SDK, set base_url to "https://router.requesty.ai/v1", set your API key to your Requesty key, and set the model to "fireworks/nemotron-3-ultra-nvfp4". The Quickstart above shows Python, JavaScript and cURL snippets.
Can I run nemotron-3-ultra-nvfp4 through Requesty?
Yes. nemotron-3-ultra-nvfp4 runs through Requesty's OpenAI-compatible API, served from Fireworks AI. You do not host the model yourself: point base_url at Requesty, set the model to "fireworks/nemotron-3-ultra-nvfp4", and requests are routed to the upstream provider with automatic failover. The same key gives you 400+ other models too.

Access nemotron-3-ultra-nvfp4 through Requesty

One API key, 400+ models, OpenAI-compatible. No markup on provider prices, automatic failover, and smart caching built-in.

Get started freeAll Fireworks AI models