# NVIDIA Nemotron 3 Ultra

Model tracker · checked 2026-10-14 · https://dbggr.dev/models/nvidia-nemotron-3-ultra/

NVIDIA Nemotron 3 Ultra is an everyday open-weight model from NVIDIA, released 4 June 2026. At DeepInfra it costs $0.50 per million input tokens and $2.20 per million output tokens, with a 256K-token context window. 4 providers serve it, and DeepInfra is the cheapest.

- Made by: NVIDIA Corporation
- Released: 2026-06-04
- Status: Generally available
- Tier: Everyday
- Weights: Open
- Context window: 256,000 tokens
- Max output: 16,384 tokens
- Input: text
- API id: nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B
- Docs: https://deepinfra.com/pricing

## NVIDIA Nemotron 3 Ultra price at 4 providers

US dollars per million tokens. Blended is 3 input tokens for every output token, one number to compare on. The cheapest blended price is highlighted.

| Provider | Input | Output | Cached input | Blended | Batch | Checked |
|---|---|---|---|---|---|---|
| DeepInfra | $0.50 | $2.20 | $0.10 | $0.925 | – | [2026-10-14](https://deepinfra.com/pricing) |
| Baseten | $0.60 | $2.40 | $0.12 | $1.05 | – | [2026-10-14](https://www.baseten.co/pricing/) |
| Fireworks AI | $0.60 | $2.40 | $0.12 | $1.05 | 50% off | [2026-10-14](https://docs.fireworks.ai/serverless/pricing) |
| Nebius Group N.V. | $1.00 | $3.00 | – | $1.50 | – | [2026-10-14](https://tokenfactory.nebius.com/model-catalog.md) |

- Fireworks AI: Priority tier: $0.75 / $3.00.
- Nebius Group N.V.: No cached-input or batch price published.

## NVIDIA Nemotron 3 Ultra benchmarks

| Benchmark | Score | What it measures | Reported by |
|---|---|---|---|
| LMArena text score | 1426 | Elo-style score from people voting between answers | [Independent](https://arena.ai/leaderboard/text) |

Scores reported by a model's maker aren't independently checked.

## Models like NVIDIA Nemotron 3 Ultra

| Model | Input | Output | Context | Weights |
|---|---|---|---|---|
| NVIDIA Nemotron 3 Ultra | $0.50 | $2.20 | 256K | Open |
| [Mistral Large 3](https://dbggr.dev/models/mistral-large-3/) | $0.50 | $1.50 | 256K | Open |
| [Qwen3.7 Plus](https://dbggr.dev/models/qwen3-7-plus/) | $0.40 | $1.60 | 1M | Closed |
| [Step 5 Preview](https://dbggr.dev/models/step-5-preview/) | $1.00 | $2.70 | 1M | Closed |
| [MiniMax M3](https://dbggr.dev/models/minimax-m3/) | $0.30 | $1.20 | 1M | Open |
| [DeepSeek V4 Pro](https://dbggr.dev/models/deepseek-v4-pro/) | $1.32 | $3.96 | 1M | Open |

## FAQ

### How much does NVIDIA Nemotron 3 Ultra cost?

NVIDIA Nemotron 3 Ultra costs $0.50 per million input tokens and $2.20 per million output tokens at DeepInfra, and $0.10 per million cached input tokens. Prices are list prices in US dollars, last checked 14 October 2026.

### What is NVIDIA Nemotron 3 Ultra's context window?

NVIDIA Nemotron 3 Ultra takes up to 256,000 tokens in one request and writes up to 16,384 tokens in one response.

### Is NVIDIA Nemotron 3 Ultra open weight?

Yes. NVIDIA publishes the weights, so you can run NVIDIA Nemotron 3 Ultra yourself or at any host that serves it.

### Where can I use NVIDIA Nemotron 3 Ultra?

NVIDIA Nemotron 3 Ultra is available from DeepInfra, Baseten, Fireworks AI and Nebius Group N.V.. At DeepInfra, call it as `nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B`.