What's the actual difference between an open-source LLM like DeepSeek and a closed one like ChatGPT?

The core difference is control and access. With a closed model like ChatGPT, you're using a service someone else runs. You send a message, their servers process it, and you get a reply. With an open-source model like DeepSeek, you download the actual program files and can run them on your own computer. It's like the difference between streaming a movie on Netflix versus owning the DVD. One you just watch, the other you can take apart, copy, and play on any device you own. This plays out in a few practical ways. Closed models are usually easier to start with. You open a browser, log in, and you're chatting. Open models require some setup. You'll need decent hardware — a strong graphics card helps — and a bit of patience to get everything running. But once it's running, you have real privacy. Your conversations never leave your machine. For a business handling sensitive customer data, that's a big deal. Cost is another factor. ChatGPT has a free tier, but heavy use or access to the smartest models costs $20 a month. Running DeepSeek on your own hardware costs whatever your electricity bill is. But here's the catch: that hardware isn't cheap. A setup that runs these models smoothly can run you a few thousand dollars upfront. You're trading a subscription for an investment. A tip I've learned the hard way: don't assume 'open-source' means 'just as good.' The community versions of these models are often smaller and less polished than what the big companies serve up. But they're improving incredibly fast. The gap is closing. Try a closed model first to learn what's possible. Then, if you need privacy or want to tinker, explore the open-source options. You'll have a much better idea of what you're looking for.