Deploy DeepSeek on a Cloud GPU

Self-host DeepSeek V3 or R1 with vLLM on an A100 or H100. Full CLI setup, serve, and test.

Run DeepSeek V3 (685B MoE) or DeepSeek R1 on your own GPU. DeepSeek models use mixture-of-experts — only ~37B parameters are active per token, so they fit on fewer GPUs than the total parameter count suggests.

GPU requirements

Model	GPU	VRAM needed	Approx. cost
DeepSeek R1 Distill 8B	RTX 4090 (24 GB)	~16 GB	~$0.35/hr
DeepSeek R1 Distill 70B	A100 80 GB	~70 GB	~$1.60/hr
DeepSeek V3 / R1 full (FP8)	4x H100 80 GB	~50 GB each	~$10.00/hr

---

Deploy DeepSeek R1 Distill 8B (RTX 4090)

runcrate instances create --name deepseek-8b --gpu RTX4090
runcrate instances status deepseek-8b

runcrate ssh deepseek-8b -- "pip install vllm"

runcrate ssh deepseek-8b -- "nohup python -m vllm.entrypoints.openai.api_server \
  --model deepseek-ai/DeepSeek-R1-Distill-Llama-8B \
  --max-model-len 8192 \
  --port 8000 --host 0.0.0.0 \
  > /root/vllm.log 2>&1 &"

Deploy DeepSeek R1 Distill 70B (A100)

runcrate instances create --name deepseek-70b --gpu A100
runcrate ssh deepseek-70b -- "pip install vllm"

runcrate ssh deepseek-70b -- "nohup python -m vllm.entrypoints.openai.api_server \
  --model deepseek-ai/DeepSeek-R1-Distill-Llama-70B \
  --max-model-len 8192 \
  --port 8000 --host 0.0.0.0 \
  > /root/vllm.log 2>&1 &"

Deploy DeepSeek V3 full (4x H100)

runcrate instances create --name deepseek-v3 --gpu H100 --gpu-count 4
runcrate ssh deepseek-v3 -- "pip install vllm"

runcrate ssh deepseek-v3 -- "nohup python -m vllm.entrypoints.openai.api_server \
  --model deepseek-ai/DeepSeek-V3 \
  --tensor-parallel-size 4 \
  --max-model-len 16384 \
  --trust-remote-code \
  --port 8000 --host 0.0.0.0 \
  > /root/vllm.log 2>&1 &"

Test the endpoint

runcrate instances info deepseek-8b

curl http://<INSTANCE_IP>:8000/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-ai/DeepSeek-R1-Distill-Llama-8B",
    "messages": [{"role": "user", "content": "Explain mixture-of-experts in two sentences."}],
    "max_tokens": 256
  }'

Monitoring

runcrate ssh deepseek-8b -- nvidia-smi
runcrate ssh deepseek-8b -- "tail -50 /root/vllm.log"

Cleanup

runcrate instances delete deepseek-8b
runcrate instances delete deepseek-70b
runcrate instances delete deepseek-v3