NVIDIA Nemotron 3 Ultra API pricing
Updated
Among the 10 providers we checked (prices read 2026-11-14 to 2026-11-16), the lowest blended standard price for NVIDIA Nemotron 3 Ultra is CoreWeave Forge at USD 0.50 per 1M input tokens and USD 2.15 per 1M output tokens. The median across 10 standard prices is USD 0.55 / USD 2.40. Each price is read from the provider's own pricing page.
Does NVIDIA sell NVIDIA Nemotron 3 Ultra directly?
NVIDIA publishes no price for NVIDIA Nemotron 3 Ultra on its own API that we could find, so this page compares hosts only.
NVIDIA Nemotron 3 Ultra model facts
Facts from NVIDIA's model card for nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16 and the Hugging Face model API, read 2026-11-16. BroadNet did not verify them, and hosts may serve the model at a different precision. BroadRouter is BroadNet's site.
| Fact | As stated | Source |
|---|---|---|
| Open weights | Yes. The weights are in a public Hugging Face repository (nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16) with no access gate. | Hugging Face model API, checked 2026-11-16Quote“"private": false, "gated": false, 225 .safetensors files listed” |
| Licence | OpenMDW License Agreement, version 1.1 | NVIDIA model card, checked 2026-11-16Quote“License | OpenMDW License Agreement, version 1.1” |
| Total parameters | 550B | NVIDIA model card, checked 2026-11-16Quote“Number of model parameters: 550B Total / 55B Active” |
| Active parameters per token | 55B | NVIDIA model card, checked 2026-11-16Quote“Number of model parameters: 550B Total / 55B Active” |
| Context length | Up to 1M tokens | NVIDIA model card, checked 2026-11-16Quote“Context Length | Up to 1M tokens” |
| Input and output | Text in; text out | NVIDIA model card, checked 2026-11-16Quote“Input Type(s): Text … Output Type(s): Text” |
| Tool calling | Listed among its uses | NVIDIA model card, checked 2026-11-16Quote“complex agentic workflows, long-context analysis, tool use” |
| Release date | 2026-06-04 | NVIDIA model card, checked 2026-11-16Quote“Release Date | June 4, 2026” |
| First published on Hugging Face (not necessarily the release date) | 2026-06-03 | Hugging Face model API, checked 2026-11-16Quote“"createdAt": "2026-06-03T14:50:04.000Z"” |
NVIDIA Nemotron 3 Ultra price per provider
USD per 1M tokens. Standard prices are ranked on a blend of 3 input tokens to 1 output token, cheapest first; the dearest standard price is 37% above the cheapest. Free rows, the maker's serving version and other service tiers follow, tagged and unranked.
| # | Provider | Input | Cached input | Output | Blended 3:1 | Context | Precision | Source |
|---|---|---|---|---|---|---|---|---|
| 1 | CoreWeave Forge cheapestnvidia/NVIDIA-Nemotron-3-Ultra-550B-A55BNotes and source quote“NVIDIA Nemotron 3 Ultra $0.50 $2.15 $0.10” | USD 0.50 | USD 0.10 | USD 2.15 | USD 0.9125 | 262,144 | not stated | CoreWeave Forge page checked 2026-11-15 re-checked 2026-11-16 |
| 2 | DeepInfranvidia/NVIDIA-Nemotron-3-Ultra-550B-A55BNotes and source quote“$0.50 in $2.20 out $0.10 cached / 1M tokens” | USD 0.50 | USD 0.10 | USD 2.20 | USD 0.925 | 262,144 | FP4 | DeepInfra page checked 2026-11-14 re-checked 2026-11-16 |
| 3 | DoublewordNotes and source quoteRealtime priority. “Nemotron 3 Ultra 550B A55B $0.50 / $0.13 / $2.20 $0.38 / $0.10 / $1.65 $0.25 / $0.06 / $1.10” | USD 0.50 | USD 0.13 | USD 2.20 | USD 0.925 | not stated | not stated | Doubleword page checked 2026-11-16 |
| 4 | Hyperbolicnemotron-3-ultra-550b-a55bNotes and source quoteContext shown on page as "262K" (rounded); stored as the decimal value. “Nemotron 3 Ultra nemotron-3-ultra-550b-a55b NVIDIA · 262K context · Text $0.50 $2.20 $0.10” | USD 0.50 | USD 0.10 | USD 2.20 | USD 0.925 | 262,000 | not stated | Hyperbolic page checked 2026-11-14 re-checked 2026-11-16 |
| 5 | ParasailNotes and source quote“Nemotron 3 Ultra 550B (NVFP4) $0.50 $2.50 $0.10” | USD 0.50 | USD 0.10 | USD 2.50 | USD 1 | not stated | NVFP4 | Parasail page checked 2026-11-14 re-checked 2026-11-16 |
| 6 | BasetenNotes and source quote“NVIDIA Nemotron 3 Ultra $0.60 $0.12 $2.40” | USD 0.60 | USD 0.12 | USD 2.40 | USD 1.05 | not stated | not stated | Baseten page checked 2026-11-14 re-checked 2026-11-16 |
| 7 | Fireworks AIaccounts/fireworks/models/nemotron-3-ultra-nvfp4Notes and source quoteListed as Preview. Priced in the serverless pricing table; not shown in the serverless model library filter on 2026-11-14. “NVIDIA Nemotron 3 Ultra (Preview) | $0.60 / $0.12 / $2.40 | $0.75 / $0.15 / $3.00” | USD 0.60 | USD 0.12 | USD 2.40 | USD 1.05 | not stated | NVFP4 | Fireworks AI page checked 2026-11-14 re-checked 2026-11-16 |
| 8 | Microsoft AzureNotes and source quoteFoundry Models serverless, served by Fireworks on Azure, Data Zone deployment. Name abbreviation "3U" read as Nemotron 3 Ultra. “FW NVIDIA Nemotron 3U Data Zone Input: $0.66/1M Tokens | Cached Input: $0.13/1M Tokens | Output: $2.64/1M Tokens” | USD 0.66 | USD 0.13 | USD 2.64 | USD 1.155 | not stated | not stated | Microsoft Azure page checked 2026-11-14 re-checked 2026-11-16 |
| 9 | Venicenvidia-nemotron-3-ultra-550b-a55bNotes and source quote“| NVIDIA Nemotron 3 Ultra | `nvidia-nemotron-3-ultra-550b-a55b` | $0.625 | $3.125 | $0.1875 | — | Private |” | USD 0.625 | USD 0.1875 | USD 3.125 | USD 1.25 | 256,000 | FP8 | Venice page checked 2026-11-15 re-checked 2026-11-16 |
| 10 | GMI Cloudnvidia/nemotron-3-ultra-550b-a55bNotes and source quote“Price (per 1M tokens) Input $0.80 Output $2.60 Cache Read $0.10 Cache Write -” | USD 0.80 | USD 0.10 | USD 2.60 | USD 1.25 | 262,144 | FP8 | GMI Cloud page checked 2026-11-15 re-checked 2026-11-16 |
| — | Doubleword batchNotes and source quoteBatch (24h) priority. “Nemotron 3 Ultra 550B A55B $0.50 / $0.13 / $2.20 $0.38 / $0.10 / $1.65 $0.25 / $0.06 / $1.10” | USD 0.25 | USD 0.06 | USD 1.10 | USD 0.4625 | not stated | not stated | Doubleword page checked 2026-11-16 |
| — | Doubleword flex / spotNotes and source quoteAsync priority. “Nemotron 3 Ultra 550B A55B $0.50 / $0.13 / $2.20 $0.38 / $0.10 / $1.65 $0.25 / $0.06 / $1.10” | USD 0.38 | USD 0.10 | USD 1.65 | USD 0.6975 | not stated | not stated | Doubleword page checked 2026-11-16 |
| — | DeepInfra flex / spotnvidia/NVIDIA-Nemotron-3-Ultra-550B-A55BNotes and source quoteFlex service tier (0.8× standard). “Flex (0.8×) $0.40 $1.76 $0.08” | USD 0.40 | USD 0.08 | USD 1.76 | USD 0.74 | 262,144 | FP4 | DeepInfra page checked 2026-11-14 re-checked 2026-11-16 |
| — | Fireworks AI priorityaccounts/fireworks/models/nemotron-3-ultra-nvfp4Notes and source quotePriority mode. “NVIDIA Nemotron 3 Ultra (Preview) | Priority $0.75 / $0.15 / $3.00” | USD 0.75 | USD 0.15 | USD 3 | USD 1.313 | not stated | NVFP4 | Fireworks AI page checked 2026-11-14 re-checked 2026-11-16 |
| — | DeepInfra prioritynvidia/NVIDIA-Nemotron-3-Ultra-550B-A55BNotes and source quotePriority service tier (1.5× standard). “Priority (1.5×) $0.75 $3.30 $0.15” | USD 0.75 | USD 0.15 | USD 3.30 | USD 1.387 | 262,144 | FP4 | DeepInfra page checked 2026-11-14 re-checked 2026-11-16 |
- Ranked: standard-tier prices only, on (3 × input + output) ÷ 4. "Cached input: not published" means the provider's page shows no cache price; budget cached tokens at the input price.
- Other tiers (batch, flex / spot and priority): batch runs asynchronously within hours; flex and spot are best-effort and can be slow or refused at capacity; priority and fast cost more for reliability or speed. Shown, not ranked.
- Automated re-check 2026-11-16: a weekly job re-fetches each source page and looks for the recorded quote or prices. It is not a human review. "Re-checked" means it still found them. Rows with no re-check note show only the date the price was read.
- Precision is shown only where the provider states it; "not stated" is unknown, not full precision.
Report a wrong price or number (email to info@broadnet.ai, with this page's link filled in).
What does NVIDIA Nemotron 3 Ultra cost for a real workload?
Monthly cost over 30 days at each standard price, using the presets of the LLM API cost calculator. Providers without a cached price bill cached input at the input price.
- Support chatbot: 2,000 requests a day, 1,500 input tokens (60% cached), 300 output.
- RAG over documents: 1,000 requests a day, 8,000 input tokens (30% cached), 500 output.
- Coding agent (one developer): 1,500 requests a day, 40,000 input tokens (85% cached), 2,000 output incl. reasoning.
| Provider | Support chatbot | RAG over documents | Coding agent (one developer) |
|---|---|---|---|
| CoreWeave Forge | USD 62.10 | USD 123.45 | USD 481.50 |
| DeepInfra | USD 63.00 | USD 124.20 | USD 486.00 |
| Doubleword | USD 64.62 | USD 126.36 | USD 531.90 |
| Hyperbolic | USD 63.00 | USD 124.20 | USD 486.00 |
| Parasail | USD 68.40 | USD 128.70 | USD 513.00 |
| Baseten | USD 71.28 | USD 145.44 | USD 561.60 |
| Fireworks AI | USD 71.28 | USD 145.44 | USD 561.60 |
| Microsoft Azure | USD 78.30 | USD 159.84 | USD 614.70 |
| Venice | USD 88.88 | USD 165.38 | USD 736.88 |
| GMI Cloud | USD 81.00 | USD 180.60 | USD 603.00 |
For the support chatbot, the cheapest price (CoreWeave Forge, USD 62.10 a month) saves USD 7.74 against the median (USD 69.84). For the coding agent the gap is USD 65.25 a month (12% lower).
Which NVIDIA Nemotron 3 Ultra provider is cheapest for your traffic?
CoreWeave Forge has the lowest input price (USD 0.50) and the lowest output price (USD 2.15), so it is the cheapest standard price at every mix of input and output tokens, before cache discounts.
| Provider | No cache hits | 50% cached | 90% cached |
|---|---|---|---|
| CoreWeave Forge | USD 0.9125 cheapest | USD 0.7625 cheapest | USD 0.6425 cheapest |
| DeepInfra | USD 0.925 | USD 0.775 | USD 0.655 |
| Doubleword | USD 0.925 | USD 0.7863 | USD 0.6753 |
| Hyperbolic | USD 0.925 | USD 0.775 | USD 0.655 |
| Parasail | USD 1 | USD 0.85 | USD 0.73 |
| Baseten | USD 1.05 | USD 0.87 | USD 0.726 |
USD per 1M tokens, 3 input tokens to 1 output token, each provider's cheapest standard row; 4 dearer providers are in the price table above. A provider without a cached price bills cached input at the input price. Any charge for writing to the cache, which some providers add, is not included.
All 10 providers publish a cached-input price below the input price: CoreWeave Forge (80% off), DeepInfra (80% off), Doubleword (74% off), Hyperbolic (80% off), Parasail (80% off) and 5 more.
Context windows differ: 262,144 tokens at CoreWeave Forge, DeepInfra and GMI Cloud; 262,000 tokens at Hyperbolic; 256,000 tokens at Venice. Doubleword, Parasail, Baseten, Fireworks AI and Microsoft Azure do not state one. The cheapest row is not always the one that fits a long prompt.
What does NVIDIA Nemotron 3 Ultra cost in other languages?
The same text needs a different number of tokens in each language. With the Nemotron 3 Ultra tokenizer, sending 1M characters costs this much at the cheapest standard input price we found for NVIDIA Nemotron 3 Ultra:
| Content | Input per 1M characters | Cheapest provider |
|---|---|---|
| English | USD 0.109 | CoreWeave Forge |
| Arabic MSA | USD 0.149 | CoreWeave Forge |
| Hindi | USD 0.190 | CoreWeave Forge |
| Chinese | USD 0.451 | CoreWeave Forge |
| Code | USD 0.135 | CoreWeave Forge |
From BroadRouter's real cost by language dataset, which covers 15 languages and content types for every model we could match to a measured tokenizer.
Published benchmark results
BroadNet did not run these benchmarks; results are as published by the NVIDIA model card and Epoch AI and may use different settings. BroadRouter is BroadNet's site.
| Benchmark | Result | Setting or note | Source |
|---|---|---|---|
| GPQA (no tools) | 87.0 | not stated | NVIDIA model card, checked 2026-11-16 |
| SWE-Bench Verified | 70.7 | not stated | NVIDIA model card, checked 2026-11-16 |
| LiveCodeBench (v6) | 89.0 | not stated | NVIDIA model card, checked 2026-11-16 |
| Terminal Bench 2.1 | 56.4 | not stated | NVIDIA model card, checked 2026-11-16 |
| GPQA Diamond | 0.854 | Setting as listed by Epoch (no suffix); Epoch run nemotron-3-ultra, mean score on a 0 to 1 scale | Epoch AI, checked 2026-11-16 |
Epoch AI rows: Epoch AI, Capabilities & benchmarking, licensed CC BY 4.0, shown as the mean score on a 0 to 1 scale, rounded to three decimals (hover a value for the exact figure in Epoch's data file).
Estimate your own bill
- LLM API cost calculator with NVIDIA Nemotron 3 Ultra preloaded: your traffic and cache rate, against any of 431 models.
- Nemotron 3 token counter: count the tokens in your real prompts. Tokenizers differ, so the same text costs a different number of tokens on different models.
- Prompt caching calculator: what a higher cache-hit rate saves.
How does NVIDIA Nemotron 3 Ultra compare with other NVIDIA models?
The NVIDIA models whose cheapest standard price is closest to NVIDIA Nemotron 3 Ultra's (USD 0.9125 blended):
| Model | Cheapest input | Cheapest output | Max context | Providers |
|---|---|---|---|---|
| NVIDIA Nemotron 3 Super 120B A12B | USD 0.075 | USD 0.40 | 262,144 | 3 |
| NVIDIA Nemotron 3.5 Super VL | USD 0.07 | USD 0.37 | 1,000,000 | 7 |
| NVIDIA Nemotron 3 Nano 30B A3B | USD 0.05 | USD 0.20 | 262,144 | 3 |
| NVIDIA Nemotron 3.5 Lightning 30B A3B | USD 0.06 | USD 0.16 | 1,048,576 | 5 |
Priced closest from other makers: DeepSeek R1 (DeepSeek, USD 0.92) · Kimi K2 Thinking (Moonshot AI, USD 0.90) · Kimi K2.5 (Moonshot AI, USD 0.90).
Compare NVIDIA Nemotron 3 Ultra with
- GLM 5.3 vs Nemotron 3 Ultra: API price side by side
Also compared: vs GLM 5.3.
Every NVIDIA Nemotron 3 Ultra price comparison · All model-vs-model price comparisons
All 431 models: API prices compared · Cheapest LLM APIs right now · LLM API price index (open data) · Download all prices (CSV, JSON) · Price change log
Questions
How much does the NVIDIA Nemotron 3 Ultra API cost per token?
From USD 0.50 per 1M input tokens (USD 0.0000005 per token) and USD 2.15 per 1M output tokens (USD 0.00000215 per token) at the lowest standard prices we found. The median of 10 standard prices is USD 0.55 input and USD 2.40 output. Each price was read from the provider's own pricing page; the source is linked in its row.
Which provider is cheapest for NVIDIA Nemotron 3 Ultra?
Among the 10 providers we checked (prices read 2026-11-14 to 2026-11-16), the lowest is CoreWeave Forge, ranked on a blend of 3 input tokens to 1 output token (USD 0.9125 per 1M). For input alone the lowest price is USD 0.50 (CoreWeave Forge and 4 others, tied); for output alone, USD 2.15 (CoreWeave Forge). Batch, flex, priority and fast tiers are shown in the table but not ranked.
Does NVIDIA sell NVIDIA Nemotron 3 Ultra on its own API?
NVIDIA publishes no price for NVIDIA Nemotron 3 Ultra on its own API that we could find, so there is no maker price to compare. The lowest-priced host we checked is CoreWeave Forge at USD 0.50 / USD 2.15 per 1M tokens.
Does NVIDIA Nemotron 3 Ultra have cached input pricing?
All 10 standard prices include one. The lowest is USD 0.10 per 1M at CoreWeave Forge; the median is USD 0.11. Providers without one bill cached input at the input price. What cached input means.
What is the NVIDIA Nemotron 3 Ultra context window?
Up to 262,144 tokens. 2 standard prices come with less, down to 256,000 tokens (Venice and Hyperbolic). Check the context column before you pick the cheapest row for long prompts. Context window calculator.
How much does a coding agent cost per month on NVIDIA Nemotron 3 Ultra?
For 1,500 requests a day, 40,000 input tokens (85% cached), 2,000 output incl. reasoning, over 30 days: USD 481.50 a month at CoreWeave Forge, USD 546.75 at the median price and USD 736.88 at Venice. Try your own numbers.
Is there a batch, flex or priority price for NVIDIA Nemotron 3 Ultra?
Yes: batch from Doubleword at USD 0.25 / USD 1.10; flex / spot from Doubleword at USD 0.38 / USD 1.65; priority from Fireworks AI at USD 0.75 / USD 3 per 1M input / output tokens. These tiers trade speed or availability for price (batch runs within hours, flex and spot can be slow or refused at capacity, priority and fast cost more for speed or reliability), so they are shown but not ranked.
Is NVIDIA Nemotron 3 Ultra free to use through an API?
No provider in our data lists a free price for NVIDIA Nemotron 3 Ultra as of 2026-11-16. Free tiers and credits by provider (checked 2026-11-14) lists general offers that do not name a model; all free LLM API models are on one page.
Is NVIDIA Nemotron 3 Ultra open source, with open weights, and what is its licence?
NVIDIA publishes the weights on Hugging Face as nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16, in a public repository with no access gate. The licence, as the NVIDIA model card states it: OpenMDW License Agreement, version 1.1. Open weights means you can download and run the model yourself; what you may do with it, including commercial use, is set by that licence, so read it in full (checked 2026-11-16). These facts come from NVIDIA's card and the Hugging Face model API; BroadNet did not verify them, and hosts may serve the model at a different precision.
How many parameters does NVIDIA Nemotron 3 Ultra have?
Total parameters, as the NVIDIA model card states them: 550B. Active parameters per token: 55B. (checked 2026-11-16; BroadNet did not verify these figures.)
Does NVIDIA Nemotron 3 Ultra support tool calling?
Listed among its uses, according to the NVIDIA model card: “complex agentic workflows, long-context analysis, tool use” (checked 2026-11-16). Whether the tools parameter works through an API also depends on the provider serving it, so check that provider's docs. BroadNet did not test it.
How do I call NVIDIA Nemotron 3 Ultra with the OpenAI SDK?
BroadNet does not serve NVIDIA Nemotron 3 Ultra. Use the OpenAI-compatible base URL of a provider that lists it: create the client with OpenAI(base_url=..., api_key=...) and pass that provider's model ID. Documented base URLs for providers on this page, in alphabetical order: DeepInfra (https://api.deepinfra.com/v1/openai, model nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B) and Fireworks AI (https://api.fireworks.ai/inference/v1, model accounts/fireworks/models/nemotron-3-ultra-nvfp4). Each provider page quotes the provider's docs with a check date.
How does NVIDIA Nemotron 3 Ultra score on benchmarks?
As published, the NVIDIA model card reports 87.0 on GPQA (no tools) and 70.7 on SWE-Bench Verified; Epoch AI measured 0.854 on GPQA Diamond in its own run, as a mean score on a 0 to 1 scale (checked 2026-11-16). BroadNet did not run these benchmarks, and sources may use different settings, so the figures are not directly comparable. Every published result we found, with its setting and source.
Read as MarkdownAsk ChatGPTAsk ClaudeAsk Perplexity
Sources
- Microsoft Azure: pricing (checked 2026-11-14)
- Baseten: pricing (checked 2026-11-14)
- CoreWeave Forge: pricing (checked 2026-11-15)
- DeepInfra: pricing (checked 2026-11-14)
- Doubleword: pricing (checked 2026-11-16)
- Fireworks AI: pricing (checked 2026-11-14)
- GMI Cloud: pricing (checked 2026-11-15)
- Hyperbolic: pricing (checked 2026-11-14)
- Parasail: pricing (checked 2026-11-14)
- Venice: pricing (checked 2026-11-15)