Nemotron 3 Ultra
Summary
Nemotron 3 Ultra is an open-weight language model from NVIDIA, released on 4 Jun 2026. It accepts text and generates text, with a 1M-token context window and up to 65K output tokens per response. As of 18 Nov 2026, the lowest listed API price is $0.50 per million input tokens and $2.20 per million output tokens (OpenRouter), across 5 providers we track.
Largest Nemotron 3 model for maximum open-weight reasoning and agent accuracy
Last verified 18 Nov 2026. Every figure on this page links to its source.
Specifications
| Developer | NVIDIA |
|---|---|
| Release date | 4 Jun 2026 [models.dev] |
| Status | Generally available [models.dev] |
| Weights | Open weights [Hugging Face] |
| License | openmdw-1.1 [Hugging Face] |
| Parameters | 561B [Hugging Face] |
| Context window | 1M (1,000,000 tokens) [models.dev] |
| Max output | 65K tokens [models.dev] |
| Knowledge cutoff | Sept 2025 [models.dev] |
| Input | Text [models.dev] |
| Output | Text [models.dev] |
| Reasoning mode | Yes [models.dev] |
| Tool calling | Yes [models.dev] |
| Structured output | Yes [models.dev] |
| Hugging Face | nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16 [Hugging Face] |
Nemotron 3 Ultra API pricing by provider
USD per million tokens. Sorted by input price.
| Provider | Input | Output | Cached input | Context | Source |
|---|---|---|---|---|---|
| NVIDIA NIM nvidia/nemotron-3-ultra-550b-a55b | Free | Free | – | 1M | models.dev, 18 Nov 2026 |
| OpenRouter nvidia/nemotron-3-ultra-550b-a55b | $0.50 | $2.20 | $0.10 | 262K | OpenRouter, 18 Nov 2026 |
| Baseten nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B | $0.60 | $2.40 | $0.12 | 203K | models.dev, 18 Nov 2026 |
| Vercel AI Gateway nvidia/nemotron-3-ultra-550b-a55b | $0.60 | $2.40 | $0.12 | 1M | models.dev, 18 Nov 2026 |
| Nebius AI Studio nvidia/Nemotron-3-Ultra-550b-a55b | $1 | $3 | $0.12 | 1M | models.dev, 18 Nov 2026 |
Running Nemotron 3 Ultra locally
Estimated memory, from the parameter count on Hugging Face. Includes about 0.8 GB of runtime overhead.
| Quantization | Weights | Total, 8K context | Total, 32K context |
|---|---|---|---|
| Q4_K_M (4-bit) | 340 GB | 341 GB | – |
| Q8_0 (8-bit) | 596 GB | 596 GB | – |
| FP16 / BF16 (16-bit) | 1121 GB | 1122 GB | – |
At 4-bit with an 8K conversation, Nemotron 3 Ultra needs about 341 GB, more than any single consumer graphics card; it needs a unified-memory machine, a multi-GPU setup or a split between graphics card and system RAM. The model's architecture is not public, so the conversation cache is not included. Check your own hardware, or see Best computer for local AI on our sister site AIPCs.
Frequently asked questions
What is the context window of Nemotron 3 Ultra?
Nemotron 3 Ultra has a context window of 1,000,000 tokens (1M) and can generate up to 65,000 tokens in a single response, according to models.dev.
How much does the Nemotron 3 Ultra API cost?
Nemotron 3 Ultra is listed from $0.50 per million input tokens and $2.20 per million output tokens (OpenRouter). Prices last verified 18 Nov 2026.
Is Nemotron 3 Ultra open source?
Nemotron 3 Ultra is an open-weight model: its weights are published on Hugging Face as nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16 under the openmdw-1.1 license. Check the license terms before commercial use.
When was Nemotron 3 Ultra released?
NVIDIA released Nemotron 3 Ultra on 4 Jun 2026, according to models.dev.
Which providers offer Nemotron 3 Ultra?
We track Nemotron 3 Ultra on 5 providers: NVIDIA NIM, OpenRouter, Baseten, Vercel AI Gateway, Nebius AI Studio.
How much VRAM does Nemotron 3 Ultra need?
About 341 GB at 4-bit (Q4_K_M) with an 8K-token conversation, and about 596 GB at 8-bit. That is more than any single consumer graphics card holds. These are estimates from the model's published parameter count.
Similar models from other labs
Comparable on output type, weights, release date, price and context window. Prices are the lowest listed, USD per million tokens.
| Model | Lab | Released | Context | Input $/1M | Output $/1M | Weights |
|---|---|---|---|---|---|---|
| GLM-5.2 | Zhipu AI (Z.ai) | 13 Jun 2026 | 1M | $0.007 | $2.40 | Open |
| Macaron V1 Venti | Mindlab | 21 Jul 2026 | 1M | $0.82 | $2.48 | Open |
| MiniMax-M3 | MiniMax | 1 Jun 2026 | 1M | $0.28 | $1.10 | Open |
| Tencent Hy4 Preview | Tencent (Hunyuan) | 28 Aug 2026 | 1M | $0.83 | $2.50 | Open |
| LongCat-2.0 | Meituan | 30 Jun 2026 | 1M | $0.30 | $1.20 | Open |
| Inkling Small | Thinkingmachines | 30 Jul 2026 | 1M | $0.45 | $1.20 | Open |
Change history
- OpenRouter price changed: input $0.60 → $0.50, output $2.40 → $2.20 per 1M tokens
- Context window changed from 203K to 1M
- Max output changed from 203K to 65K
- Max output changed from 66K to 203K
- Name changed from Nemotron 3 Ultra 550B A55B to Nemotron 3 Ultra
- Context window changed from 1M to 203K
- Context window changed from 262K to 1M
- Max output changed from 183K to 66K
- Context window changed from 1M to 262K
- Max output changed from 66K to 183K
- Max output changed from 1M to 66K
- Context window changed from 1M to 1M
- Context window changed from 1M to 1M
- Max output changed from 66K to 1M
- Name changed from Nemotron 3 Ultra to Nemotron 3 Ultra 550B A55B
- Max output changed from 65K to 66K
- Context window changed from 1M to 1M
- Max output changed from 1M to 65K
- Name changed from Nemotron 3 Ultra 550B A55B to Nemotron 3 Ultra
- Context window changed from 1M to 1M
- Name changed from Nemotron 3 Ultra to Nemotron 3 Ultra 550B A55B
- Max output changed from 65K to 1M
- Name changed from Nemotron 3 Ultra 550B A55B to Nemotron 3 Ultra
- Context window changed from 1M to 1M
- Max output changed from 1M to 65K
Sources
- https://openrouter.ai/nvidia/nemotron-3-ultra-550b-a55b
- https://github.com/sst/models.dev/blob/dev/providers/vercel/models/nvidia/nemotron-3-ultra-550b-a55b.toml
- https://huggingface.co/nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16
- https://github.com/sst/models.dev/blob/dev/providers/baseten/models/nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B.toml
- https://github.com/sst/models.dev/blob/dev/providers/nvidia/models/nvidia/nemotron-3-ultra-550b-a55b.toml
- https://github.com/sst/models.dev/blob/dev/providers/nebius/models/nvidia/Nemotron-3-Ultra-550b-a55b.toml