nemotron-3-ultra-550b-a55bNVIDIA
- api
- chat
- hosting
- US
- model lab
- NVIDIA
- weights
- open
- added
- June 2026
- model id
- nvidia/nemotron-3-ultra-550b-a55b
capabilities 2/8
2 providers serve nemotron-3-ultra-550b-a55b. This page is one of them. Compare all endpoints
- input
- free
- output
- free
- cache write
- -
- cache read
- -
worked cost
| volume | cost |
|---|---|
| 100K in + 10K out | free |
| 1M in + 100K out | free |
| 10M in + 1M out | free |
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
- median ttft
- 5.05s
- median tok/s
- 53.0
- cache hit
- 68.8%
- answered
- 86.30%
Weekly aggregates from production traffic routed through Requesty to this endpoint. A metric with no qualifying sample is left out rather than filled in. Compare against the other providers
| eval | Score as a share of 100 | /100 |
|---|---|---|
| GPQA Diamond | 86.7 | |
| τ²-Bench | 83.3 | |
| Coding Index | 49.3 | |
| SciCode | 40.3 | |
| Terminal-Bench Hard | 36.4 | |
| Humanity's Last Exam | 28.4 | |
| Intelligence Index | 23.4 | |
| 7 evals | mean | 49.7 |
Scores from Artificial Analysis and public leaderboards, normalised to 100. Bar colour is the band, not the rank: green 80 and up, blue 55 and up, amber 30 and up. Benchmarks measure narrow skills, so test on your own workload before committing. Released 2026-06-04.
- retention
- yes
- trains on prompts
- yes
- hosted in
- US
Terms are the provider's, not Requesty's: routing a request here puts it under them. NVIDIA Privacy Policy
Need every request to stay in one jurisdiction? EU hosting routes only through EU-hosted providers.
nvidia/nemotron-3-ultra-550b-a55bthis id pins the nvidia deployment, with no routing and no failover. the managed id on the canonical page routes across all 2 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="nvidia/nemotron-3-ultra-550b-a55b", 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
NVIDIA Nemotron 3 Ultra is an open frontier reasoning and orchestration model with 55B active parameters out of 550B total (hybrid Transformer-Mamba MoE). Text input/output with up to 1M context, built for long-running agentic workflows: agent orchestration, coding agents, deep research, and complex enterprise tasks. Strong at multi-step reasoning and planning with high-throughput inference. Part of the NVIDIA Nemotron family.
questions 6
How much does nemotron-3-ultra-550b-a55b cost?
What is the context window of nemotron-3-ultra-550b-a55b?
How does nemotron-3-ultra-550b-a55b perform on benchmarks?
What can nemotron-3-ultra-550b-a55b do?
How do I use nemotron-3-ultra-550b-a55b with the OpenAI SDK?
Can I run nemotron-3-ultra-550b-a55b through Requesty?
more from nvidia 6
| endpoint | ctx | in /M |
|---|---|---|
| nemotron-3.5-lightning-30b-a3b | 1.0M | free |
| muse-glimmer-30b | 131K | free |
| nemotron-3.5-content-safety | 131K | free |
| nemotron-3-nano-omni-30b-a3b-reasoning | 131K | free |
| nemotron-3-super-120b-a12b | 1.0M | free |
| nemotron-3-nano-30b-a3b | 262K | free |
call nemotron-3-ultra-550b-a55b 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