Grok and Claude Go Head-to-Head
I was taking my afternoon walk the other day and an idea for a one panel comic strip came to me. My aim was trying to capture the strain a solid decade of living through extremely divisive US politics has taken on the American people.
I love the old 1800’s cross-hatch style of political cartoons that artist, Thomas Nast popularized. If you’ve never seen his work I’d encourage you to seek out some of his old strips. Nast’s satire was genius in its ability to capture the absurdities, inequalities, and greed of the political landscape in Gilded Age America in one frame. There are no shortage of correlations between the Gilded Age and our current time.
I decided to try a little experiment with Grok using the following prompt:
Please create a 1800's style political cartoon showing a game of tug of war. On one side is the Democratic Party mascot, the other side is the Republican Party mascot. Label the rope "The American People" and show it taught and frayed.
Disappointingly Grok produced an image too blurry to even make out. I waited and waited but the Grok just wouldn’t render a final image. It seemed perpetually stuck.
I’ve been hearing about how advanced Anthropic’s Claude was so I decided to give it a shot. It rendered an image in just a few seconds but it was far from the 1800’s “Thomas Nash style” that I was going for. The style of the image Claude generated was, strangely, somewhere between Salvadore Dali and Peppa Pig.
Surprisingly, I opened the laptop this morning and discovered Grok had continued working on my request. It produced this complete render overnight. It took a while but, in my opinion, it completely nailed the task. Compared to the drawing Claude produced I was impressed but still didn't feel like it was quite "there".
It’s only going to be this way for a short while, soon all models that have enough compute will be equally as proficient in everything. For now we live in this weird time where no single AI-model is good at everything, each has its specific weaknesses. Claude does a lot of things extremely well but clearly needs to brush up on its art lessons, Grok on its expediency.
I’ve just added to the prompt in Grok to get the drawing a little closer to my original vision:
Make each mascot overweight, wearing a three-piece suit with money spilling from their pockets.
After twenty minutes the image is still blurry so we’ll likely have to wait until tomorrow for Grok to create the next revision of the strip. The old adage, "Good things take time" still holds true I guess. There will probably be a time in the not-so-distant future that we'll feel nostalgic for waiting for anything.
All for now. Thanks so much for reading.
That's interesting. I haven't used AI a ton as we have talked about, but there is definitely a lot of options out there right now with no clear winner across all of them. I wonder if you could ask GROK to make it lower resolution if it would create the image more quickly.
I'm dabbling more and more. It's been useful for trip planning for me. I had Grok create a packing list based on the destination (Sedona) and it took weather and everything into account. The list is less than I would normally pack so I'm going to follow it and see how it goes. I'm also using it for cooking/grilling/recipes. Lower res would have probably helped the speed of render but I have a feeling the slow speed is intentional to try to motivate people to pay the premium for "super-Grok".
Yeah, that is a good point about the paying for the premium version. I've seen some impressive stuff that people are doing with different AI things, but it all feels kind of complicated to me. I really need to dig into it a bit more I think because it likely isn't going away, but who knows.
I don't use it enough to justify paying the $300/month for the most advanced version but I bet it's mind-blowing. The same holds true for the premium versions of all these AI-models I'm sure. I'm hearing rumors of Apple releasing a free version of advanced AI on their devices soon and that will obliterate many of the competitors, and trap even more consumers in their "walled garden".
I think Apple is leveraging Google Gemini for their solution. At least that is what I read at some point. I have to believe the cost will come down eventually, despite the current cost of memory, storage and overall compute power.
That will be interesting if Apple ends up using Gemini. I've heard it's very capable. I guess compute is the only thing holding many of the models back right now. I think this is what prompted the deal between Anthropic and xAI–Claude needed more computing power. Anthropic got xAI's Colossus 1 and they moved Grok to Colossus 2. It sounds like Grok is almost dead last amongst LLMs, all things considered.
Oh, that is too bad. I think we ultimately need to see some consolidation, though competition is probably good in terms of innovation.
You should try chat Gpt 5.5, I feel like Claude is totally overhyped... For the coding and Kubernetes troubleshooting GPT 5.5 was better than Claude Opus 4.7 in my experience.
That's good to know! I was actually going to try GPT and then switched to Claude because I had never used it before and was curious.
Is it justified to compare these two AI models?
I don't think so, everyone has own taste even in using AI, I myself started using AI with Meta AI but after few weeks, I switched myself to Google Gemini even I have used ChatGpt, but Google Gemini is enough for me.
I think it's great to compare and this was all for fun. Each of the LLMs have their owns strengths and weaknesses in this tiny sliver of time. It won't be that way for long though. Soon all will be equally as capable. I started with ChatGPT years ago and used to to make travel arrangements and such. When Grok launched I started using it for image generation, travel arrangements, and general research. I'm starting to dabble a little in Claude now.