Models · DeepSeek

Deploy DeepSeek models on dedicated GPUs

16 DeepSeek models ready to deploy on GPU capacity reserved for you, from 3.3B to 1.65T parameters, covering chat & reasoning and ocr & documents. Private OpenAI-compatible endpoint, European GPUs, billed per GPU-hour.

Request a deployment ↓deepseek-ai on Hugging Face ↗All makersHow dedicated works
ModelParams
DeepSeek-V4.1-FlashserverlesstrendingvisionMoE
DeepSeek		763B
DeepSeek-V4-Flash-0731serverlessMoE
DeepSeek		304B
DeepSeek-V4-FlashserverlessMoE
DeepSeek		291B
DeepSeek-R1MoE
DeepSeek		684B
DeepSeek-V4-ProMoE
DeepSeek		1.6T
DeepSeek-OCRvisionMoE
DeepSeek		3.3B
DeepSeek-OCR-2visionMoE
DeepSeek		3.4B
DeepSeek-V3.2MoE
DeepSeek		685B
DeepSeek-V3-0324MoE
DeepSeek		685B
DeepSeek-R1-0528MoE
DeepSeek		685B
DeepSeek-R1-0528-Qwen3-8B
DeepSeek		8.2B
DeepSeek-V4-Pro-0813MoE
DeepSeek		1.65T
DeepSeek-V3.2-SpecialeMoE
DeepSeek		685B
DeepSeek-V3.1MoE
DeepSeek		685B
DeepSeek-Math-V2MoE
DeepSeek		685B
DeepSeek-V3.1-TerminusMoE
DeepSeek		685B

Specs from Hugging Face, synced 2026-10-27. Machine-readable version: https://sference.com/models/deepseek-ai.md

FAQ

Which DeepSeek models can I deploy on sference?

16 today, most requested first: DeepSeek-V4.1-Flash, DeepSeek-V4-Flash-0731, DeepSeek-V4-Flash, DeepSeek-R1, DeepSeek-V4-Pro, DeepSeek-OCR, DeepSeek-OCR-2 and DeepSeek-V3.2, and 8 more. Any other DeepSeek checkpoint or a fine-tune of one can be requested with the form on this page.

How many GPUs do DeepSeek models need?

It depends on the model: the lineup runs from 3.3B to 1.65T 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 DeepSeek models commercially?

16 of the 16 listed models are under licenses that allow commercial use (MIT and Apache 2.0 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 DeepSeek model

Tell us which DeepSeek model and roughly how much traffic you expect. Fine-tunes of these models work the same way.

  1. 01
    We size it

    A performance tier, the capacity behind it, and a quote for your volume.

  2. 02
    We benchmark it

    Bring one workload with a sample of real traffic; we come back with latency and throughput numbers.

  3. 03
    It goes live

    We provision the deployment. You call it with the same API and model id as before.

No commitment. Prefer email? [email protected]