I Tested DeepSeek’s R1 and V3 Coding Skills – and we’re not All Doomed (Yet).

DeepSeek exploded into the world’s awareness this previous weekend. It stands apart for 3 powerful reasons:

1. It’s an AI chatbot from China, rather than the US

2. It’s open source.

3. It utilizes greatly less infrastructure than the huge AI tools we’ve been taking a look at.

Also: Apple scientists expose the secret sauce behind DeepSeek AI

Given the US government’s issues over TikTok and possible Chinese federal government involvement in that code, a new AI emerging from China is bound to produce attention. ZDNET’s Radhika Rajkumar did a deep dive into those problems in her post Why China’s DeepSeek could rupture our AI bubble.

In this short article, we’re avoiding politics. Instead, I’m putting both DeepSeek V3 and DeekSeek R1 through the very same set of AI coding tests I have actually tossed at 10 other large language models. According to DeepSeek itself:

Choose V3 for jobs requiring depth and precision (e.g., fixing advanced math issues, creating complex code).

Choose R1 for latency-sensitive, high-volume applications (e.g., customer support automation, basic text processing).

You can select between R1 and V3 by clicking the little button in the chat interface. If the button is blue, you’re utilizing R1.

The short response is this: impressive, but plainly not ideal. Let’s dig in.

Test 1: Writing a WordPress plugin

This test was really my very first test of ChatGPT’s shows expertise, method back in the day. My spouse required a plugin for WordPress that would assist her run an involvement device for her online group.

Also: The finest AI for coding in 2025 (and what not to utilize)

Her requirements were relatively simple. It needed to take in a list of names, one name per line. It then needed to arrange the names, and if there were duplicate names, separate them so they weren’t listed side-by-side.

I didn’t truly have time to code it for her, so I chose to offer the AI the challenge on an impulse. To my substantial surprise, it worked.

Since then, it’s been my very first test for AIs when examining their programs abilities. It needs the AI to know how to set up code for the WordPress framework and follow triggers plainly adequate to create both the user interface and program logic.

Only about half of the AIs I’ve tested can fully pass this test. Now, however, we can include one more to the winner’s circle.

DeepSeek V3 created both the user interface and program logic precisely as defined. As for DeepSeek R1, well that’s an intriguing case. The “thinking” element of R1 triggered the AI to spit out 4502 words of before sharing the code.

The UI looked different, with much broader input areas. However, both the UI and reasoning worked, so R1 also passes this test.

So far, DeepSeek V3 and R1 both passed one of four tests.

Test 2: Rewriting a string function

A user grumbled that he was not able to get in dollars and cents into a contribution entry field. As written, my code only allowed dollars. So, the test involves offering the AI the routine that I composed and asking it to rewrite it to permit both dollars and cents

Also: My preferred ChatGPT feature simply got method more effective

Usually, this results in the AI generating some routine expression validation code. DeepSeek did generate code that works, although there is room for enhancement. The code that DeepSeek V2 composed was unnecessarily long and repetitive while the reasoning before creating the code in R1 was also long.

My most significant concern is that both models of the DeepSeek recognition makes sure recognition as much as 2 decimal places, but if an extremely big number is gone into (like 0.30000000000000004), using parseFloat does not have specific rounding understanding. The R1 model likewise used JavaScript’s Number conversion without inspecting for edge case inputs. If bad data comes back from an earlier part of the routine expression or a non-string makes it into that conversion, the code would crash.

It’s odd, due to the fact that R1 did provide a really good list of tests to verify against:

So here, we have a split choice. I’m providing the indicate DeepSeek V3 since neither of these concerns its code produced would trigger the program to break when run by a user and would produce the expected outcomes. On the other hand, I need to offer a stop working to R1 since if something that’s not a string somehow enters the Number function, a crash will take place.

Which offers DeepSeek V3 two triumphes of 4, but DeepSeek R1 only one win out of four so far.

Test 3: Finding a bothersome bug

This is a test created when I had a very bothersome bug that I had problem finding. Once again, I decided to see if ChatGPT could manage it, which it did.

The challenge is that the response isn’t apparent. Actually, the obstacle is that there is an obvious response, based on the error message. But the obvious answer is the wrong response. This not just captured me, but it routinely catches some of the AIs.

Also: Are ChatGPT Plus or Pro worth it? Here’s how they compare to the complimentary variation

Solving this bug requires understanding how particular API calls within WordPress work, having the ability to see beyond the error message to the code itself, and after that understanding where to find the bug.

Both DeepSeek V3 and R1 passed this one with almost identical answers, bringing us to 3 out of four wins for V3 and two out of 4 wins for R1. That already puts DeepSeek ahead of Gemini, Copilot, Claude, and Meta.

Will DeepSeek score a crowning achievement for V3? Let’s learn.

Test 4: Writing a script

And another one bites the dust. This is a challenging test since it requires the AI to understand the interplay between 3 environments: AppleScript, the Chrome object design, and a Mac scripting tool called Keyboard Maestro.

I would have called this an unfair test because Keyboard Maestro is not a mainstream programming tool. But ChatGPT managed the test easily, understanding precisely what part of the problem is handled by each tool.

Also: How ChatGPT scanned 170k lines of code in seconds, conserving me hours of work

Unfortunately, neither DeepSeek V3 or R1 had this level of knowledge. Neither model understood that it needed to divide the job in between instructions to Keyboard Maestro and Chrome. It likewise had relatively weak knowledge of AppleScript, composing custom regimens for AppleScript that are native to the language.

Weirdly, the R1 model stopped working as well because it made a bunch of incorrect assumptions. It presumed that a front window constantly exists, which is certainly not the case. It also made the presumption that the currently front running program would constantly be Chrome, rather than clearly inspecting to see if Chrome was running.

This leaves DeepSeek V3 with 3 right tests and one stop working and DeepSeek R1 with 2 correct tests and 2 stops working.

Final ideas

I discovered that DeepSeek’s persistence on utilizing a public cloud email address like gmail.com (rather than my normal e-mail address with my corporate domain) was irritating. It likewise had a variety of responsiveness stops working that made doing these tests take longer than I would have liked.

Also: How to use ChatGPT to compose code: What it does well and what it doesn’t

I wasn’t sure I ‘d have the ability to write this post because, for most of the day, I got this mistake when attempting to sign up:

DeepSeek’s online services have recently dealt with massive harmful attacks. To ensure ongoing service, registration is briefly restricted to +86 contact number. Existing users can visit as usual. Thanks for your understanding and assistance.

Then, I got in and was able to run the tests.

DeepSeek appears to be extremely loquacious in terms of the code it creates. The AppleScript code in Test 4 was both incorrect and exceedingly long. The routine expression code in Test 2 was appropriate in V3, but it might have been composed in a manner in which made it a lot more maintainable. It failed in R1.

Also: If ChatGPT produces AI-generated code for your app, who does it truly belong to?

I’m absolutely pleased that DeepSeek V3 beat out Gemini, Copilot, and Meta. But it appears to be at the old GPT-3.5 level, which indicates there’s absolutely space for improvement. I was dissatisfied with the outcomes for the R1 model. Given the option, I ‘d still choose ChatGPT as my programs code assistant.

That stated, for a brand-new tool working on much lower facilities than the other tools, this could be an AI to view.

What do you believe? Have you attempted DeepSeek? Are you utilizing any AIs for programs assistance? Let us know in the comments below.

You can follow my everyday job updates on social media. Make sure to subscribe to my weekly upgrade newsletter, and follow me on Twitter/X at @DavidGewirtz, on Facebook at Facebook.com/ DavidGewirtz, on Instagram at Instagram.com/ DavidGewirtz, on Bluesky at @DavidGewirtz. com, and on YouTube at YouTube.com/ DavidGewirtzTV.