What's the best way to run DeepSeek rn?

I'm very new to local LLMs, I recently managed to get a few Qwen models (35B, 27B, 122B) running on my machine through llama.cpp, and I'm now starting to play around with oh-my-pi (very impressed with the results so far). I was thinking it would be cool to use deepseek as an orchestrator for hard tasks, how should I go about that? I'm asking because I read someone saying opencode has a free tier for V4 flash, which made me think that prices may vary depending on how I access it etc (I was thinking of using