OPEN MODELS / Language models (LLM)
Run deepseek-v3.1:671b on a cloud GPU
deepseek-v3.1:671b is a Language models (LLM) model by Ollama. See how much GPU memory it needs, which GPUs it fits on today and what it costs to run it yourself, billed per second.
From the project: DeepSeek-V3.1-Terminus is a hybrid model that supports both thinking mode and non-thinking mode.
GPUs it fits on today
About 467 GB of GPU memory: the model download (4-bit by default) plus working memory.
| GPU | Memory | Per hour | 10 hours | |
|---|---|---|---|---|
| 8× A100 SXM4 | 640 GB | $5.76 | $57.60 | Rent |
| 5× RTX PRO 6000 S | 480 GB | $12.15 | $121.50 | Rent |
| 8× RTX PRO 6000 Max-Q | 768 GB | $13.76 | $137.60 | Rent |
| 16× RTX 5090 | 512 GB | $14.40 | $144.00 | Rent |
| 8× RTX PRO 6000 WS | 768 GB | $18.88 | $188.80 | Rent |
Live starting prices from our catalog. You pay per second while the machine runs.
How to run deepseek-v3.1:671b
- 01
Rent a GPU
Pick one of the GPUs above. Billing is per second; stop the machine and you stop paying.
- 02
Open a terminal or Jupyter
Connect over SSH or open Jupyter in your browser. PyTorch and CUDA are ready to use.
- 03
Install Ollama and run the model
Run the commands below. Ollama downloads the model and serves an OpenAI-compatible API on port 11434.
curl -fsSL https://ollama.com/install.sh | sh
ollama run deepseek-v3.1:671bWhy run it yourself?
Your own LLM server has no per-token bill and no rate limit, and your data stays private. Pick a GPU by memory and serve the model with vLLM.
Questions
- How much GPU memory does deepseek-v3.1:671b need?
- By our estimate about 467 GB for inference (671B parameters). Higher resolution, longer context or bigger batches need more.
- How much does it cost to run deepseek-v3.1:671b?
- From $5.76 per hour on 8× A100 SXM4, billed per second. Ten hours cost about $57.60.
- Do I need a subscription?
- No. You pay only for the time your machine runs, from your USD balance. Stop it whenever you are done.
Try it on your own GPU in minutes
Pick a GPU, open Jupyter or SSH and run any open model. You pay per second and can stop anytime; no subscription, no setup fee.
- Per-second billing in USD
- PyTorch and CUDA ready in minutes
- Rent from ChatGPT or Claude too