AI Infrastructure

DeepSeek-R1 & Open-Source Reasoning: Self-Hosting with Ollama and vLLM

How to deploy DeepSeek-R1 and distilled open weights locally with vLLM or Ollama for zero-API-cost private reasoning engines.

4 min

Why DeepSeek-R1 Is a Game Changer

DeepSeek-R1 demonstrated that reinforcement learning on reasoning patterns can match closed proprietary models at a fraction of training compute. Its distilled models (14B, 32B, 70B) run efficiently on consumer and enterprise GPUs.

deploy.sh
# Run 14B distilled reasoning model locally via Ollama
ollama run deepseek-r1:14b

# Serve high-throughput vLLM OpenAI-compatible server
python3 -m vllm.entrypoints.openai.api_server \
  --model deepseek-ai/DeepSeek-R1-Distill-Qwen-14B \
  --tensor-parallel-size 2 \
  --gpu-memory-utilization 0.90 \
  --port 8000