# Deploy NVIDIA models on dedicated GPUs

17 NVIDIA models ready to deploy on GPU capacity reserved for you, from 627M to 561B parameters, covering chat & reasoning, vision & multimodal, embeddings, rerankers, speech recognition and safety & moderation. Private OpenAI-compatible endpoint, European GPUs, billed per GPU-hour.

| Model | Type | Params | Context | License | Page |
| --- | --- | --- | --- | --- | --- |
| NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16 | Chat & reasoning | 561B | 1M | OpenMDW 1.1 | https://sference.com/models/nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16.md |
| nemotron-3.5-asr-streaming-0.6b | Speech recognition | 638M | — | OpenMDW 1.1 | https://sference.com/models/nvidia/nemotron-3.5-asr-streaming-0.6b.md |
| parakeet-tdt-0.6b-v3 | Speech recognition | 627M | — | CC BY 4.0 | https://sference.com/models/nvidia/parakeet-tdt-0.6b-v3.md |
| parakeet-tdt-0.6b-v2 | Speech recognition | — | — | CC BY 4.0 | https://sference.com/models/nvidia/parakeet-tdt-0.6b-v2.md |
| NVIDIA-Nemotron-3-Nano-30B-A3B-BF16 | Chat & reasoning | 32B | 256K | NVIDIA Nemotron Open Model License | https://sference.com/models/nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16.md |
| NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16 | Chat & reasoning | 32B | 1M | OpenMDW 1.1 | https://sference.com/models/nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16.md |
| canary-1b-v2 | Speech recognition | 979M | — | CC BY 4.0 | https://sference.com/models/nvidia/canary-1b-v2.md |
| Nemotron-3-Embed-1B-BF16 | Embeddings | 1.1B | 256K | OpenMDW 1.1 | https://sference.com/models/nvidia/Nemotron-3-Embed-1B-BF16.md |
| Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16 | Vision & multimodal | 33B | 250K | NVIDIA Open Model License | https://sference.com/models/nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16.md |
| canary-qwen-2.5b | Speech recognition | 2.6B | — | CC BY 4.0 | https://sference.com/models/nvidia/canary-qwen-2.5b.md |
| NVIDIA-Nemotron-Nano-9B-v2 | Chat & reasoning | 8.9B | 128K | Custom license | https://sference.com/models/nvidia/NVIDIA-Nemotron-Nano-9B-v2.md |
| Nemotron-3-Embed-8B-BF16 | Embeddings | 8.0B | 256K | OpenMDW 1.1 | https://sference.com/models/nvidia/Nemotron-3-Embed-8B-BF16.md |
| Nemotron-Cascade-2-30B-A3B | Chat & reasoning | 32B | 256K | Custom license | https://sference.com/models/nvidia/Nemotron-Cascade-2-30B-A3B.md |
| llama-nemotron-rerank-1b-v2 | Rerankers | 1.2B | 128K | OpenMDW 1.1 | https://sference.com/models/nvidia/llama-nemotron-rerank-1b-v2.md |
| NVIDIA-Nemotron-3-Super-120B-A12B-FP8 | Chat & reasoning | 124B | 256K | NVIDIA Nemotron Open Model License | https://sference.com/models/nvidia/NVIDIA-Nemotron-3-Super-120B-A12B-FP8.md |
| Nemotron-3.5-Content-Safety | Safety & moderation | 4.3B | 128K | OpenMDW | https://sference.com/models/nvidia/Nemotron-3.5-Content-Safety.md |
| llama-nemotron-embed-1b-v2 | Embeddings | 1.2B | 128K | Custom license | https://sference.com/models/nvidia/llama-nemotron-embed-1b-v2.md |

## FAQ

### Which NVIDIA models can I deploy on sference?

17 today, most requested first: NVIDIA-Nemotron-3-Ultra-550B-A55B-BF16, nemotron-3.5-asr-streaming-0.6b, parakeet-tdt-0.6b-v3, parakeet-tdt-0.6b-v2, NVIDIA-Nemotron-3-Nano-30B-A3B-BF16, NVIDIA-Nemotron-3.5-Lightning-30B-A3B-BF16, canary-1b-v2 and Nemotron-3-Embed-1B-BF16, and 9 more. Any other NVIDIA checkpoint or a fine-tune of one can be requested with the form on this page.

### How many GPUs do NVIDIA models need?

It depends on the model: the lineup runs from 627M to 561B parameters. Each model page lists the checkpoint size and the minimum footprint on 80 GB and 141 GB GPUs. You don't pick hardware on sference: you choose a performance tier and we size the deployment.

### Can I use NVIDIA models commercially?

14 of the 17 listed models are under licenses that allow commercial use (OpenMDW 1.1, CC BY 4.0, NVIDIA Nemotron Open Model License, NVIDIA Open Model License, Custom license and OpenMDW across the lineup). Each model page summarizes its license and links the full text.

### How is a dedicated deployment priced?

Per GPU-hour on a monthly commitment, sized to your throughput and latency targets.

Request a deployment: https://sference.com/models/nvidia#request. All makers: https://sference.com/dedicated.md