I tried out DeepSeek, but I’m sticking with Gemini for now

DeepSeek is the talk of the town, so I decided to test it out to see how it compares with my favorite AI chatbot, Gemini. I’ve been using Gemini for a while now and am very familiar with its pros and cons. And while I’m a Gemini Advanced user, I switched to the free version for this comparison just to level the playing field.

The comparison is based on my specific use case. I primarily use AI chatbots for help with content ideas, fixing typos, brainstorming titles, and other tasks related to my work. I also use them to research topics I’m interested in, whether it’s tech, design, travel, finance, or anything else I’m currently obsessed with.

DeepSeek vs. Gemini: How do they compare?

I started by testing how well DeepSeek helps me with outlines for articles I’m working on. I used the same prompt for both DeepSeek and Gemini to compare the differences.

My first impression is that DeepSeek is very good. However, I wouldn’t say it’s better than Gemini overall. The two outlines I received were equally good but structured a bit differently. I kind of prefer how DeepSeek presents the data, using bullets and sub-bullets that make it easier to read, although this is just a personal preference.

DeepSeek also did a better job proofreading a post riddled with typos. It found 20 of them, while Gemini missed a few and only found 15. But when brainstorming titles for a specific post, both were on par.

I then moved on to the next topic. I’m a finance nerd and love following the stock market, so I asked both of them to give me an overview of yesterday’s stock market. Here’s where I came across the first big difference: Gemini provided a decent overview without detailed data on how much the major indexes went down, while DeepSeek only provided instructions on where I could find the data. DeepSeek’s cut-off date is October 2023, so it can’t help with recent events.

I then asked both chatbots for help picking a new smartphone. When asked for a list of the best phones available, the results contained more or less the same models, but the way the data was presented was different. Gemini broke things down into high-end, mid-range, and foldable models, while DeepSeek suggested target users for each model. I prefer DeepSeek’s approach, although the difference is minimal. Both of them suggested older models, though, as neither of them had access to up-to-date data for this particular question.

The story was the same with the rest of my questions. The answers were similar, but DeepSeek structured them better, in my opinion. It also went into more detail and was more direct thanks to the bullet points it used, while Gemini was a bit chattier overall.

However, Gemini is much, much faster. This is especially noticeable with longer, more detailed answers. Gemini also works more consistently, while DeepSeek constantly gave me a “Server is busy” answer, sometimes even 10 times in a row. I think this is temporary, though, as everyone is testing out the chatbot due to the recent hype surrounding it.

Here’s why I’m not switching, yet

DeepSeek is good. It’s better than Gemini in some cases, as it structures data better and provides more detail. However, it’s no match for Google’s chatbot regarding speed and when asking about more recent events like the stock market — even though this can be hit-and-miss.

The reason why I’m not switching is because DeepSeek doesn’t offer any significant benefits for my use case. I’m happy with Gemini’s responses, and even though DeepSeek does a better job in certain cases, the difference is minimal. The difference in speed and general performance is more significant to me, which is one reason I’m not ready to ditch Gemini yet.

Then there’s Gemini Live, which lets me have a conversation with the chatbot that speeds up the process. A similar feature isn’t available on DeepSeek, although I assume it’s in the works.

Gemini also includes Google’s Imagen 3 image generator, allowing me to create images with text prompts. The web version of DeepSeek doesn’t have this capability, although you can run the Janus Pro model locally to create images. It’s not something I want to do at the moment, but it’s there if you want it. We’re testing it out to see how it compares, though, so stay tuned for that.

Another thing to note is that while the free versions of Gemini and DeepSeek are comparable, Gemini also has a paid tier that I use, and it’s far more capable. I prompted it with the same questions as the free versions, and the answers were much better. Gemini Advanced also has the Gems feature I use daily, as well as Deep Research, which has its issues but shows promise.

So with all that in mind, I’m sticking with Gemini for now. But DeepSeek is impressive, especially considering it was developed in only two months — at least that’s what the company behind it claims. I’m looking forward to seeing what it can do in the future, and if it ever offers significantly more than Gemini overall, I’ll be more open to making the switch.