nemotron-3-ultra-nvfp4Fireworks AI
- api
- chat
- hosting
- US
- model lab
- NVIDIA
- weights
- open
- added
- June 2026
- model id
- fireworks/nemotron-3-ultra-nvfp4
capabilities 3/8
- input
- $0.60
- output
- $2.40
- cache write
- -
- cache read
- $0.12
worked cost
| volume | cost |
|---|---|
| 100K in + 10K out | $0.0840 |
| 1M in + 100K out | $0.84 |
| 10M in + 1M out | $8.40 |
Requesty charges what the upstream provider charges, with no markup and no per-request fee. Prompt caching and smart routing cut the effective cost further on repeated context. Gateway pricing
no measured traffic for this endpoint yet. pricing above is the provider's own
no published scores for this exact variant. region deployments and highspeed tiers usually share the base model's results
- retention
- none
- trains on prompts
- no
- hosted in
- US
Terms are the provider's, not Requesty's: routing a request here puts it under them. Fireworks Privacy Policy
Need every request to stay in one jurisdiction? EU hosting routes only through EU-hosted providers.
fireworks/nemotron-3-ultra-nvfp4this id pins the fireworks ai deployment, with no routing and no failover. the managed id on the canonical page routes across all 1 providers instead
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)
Change the base url to https://router.requesty.ai/v1, use your Requesty key, set the model to the id above. Existing OpenAI SDK code needs no other edit, and the same key reaches every other model in the catalog. Browse all models
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.
questions 5
How much does nemotron-3-ultra-nvfp4 cost?
What is the context window of nemotron-3-ultra-nvfp4?
What can nemotron-3-ultra-nvfp4 do?
How do I use nemotron-3-ultra-nvfp4 with the OpenAI SDK?
Can I run nemotron-3-ultra-nvfp4 through Requesty?
more from fireworks ai 8
| endpoint | ctx | in /M |
|---|---|---|
| deepseek-v4.1-flash | 1.0M | $0.22 |
| deepseek-v4-flash-vision-exp | 1.0M | $0.22 |
| glm-5.3-flash | 1.0M | $0.15 |
| glm-5.3 | 1.0M | $1.40 |
| nemotron-lightning-3.5-30b-a3b | 262K | $0.05 |
| deepseek-v4-pro-0813 | 1M | $1.32 |
| muse-glimmer-30b | 131K | $0.35 |
| qwen3.8-max | 1M | $2.00 |
call nemotron-3-ultra-nvfp4 through one endpoint
One key for this endpoint and 600+ other models. No markup on provider prices, automatic failover when a provider degrades, and prompt caching built in. Methodology