ChatGPT vs Claude vs Gemini: The Best AI Model for Each Use Case in 2025

Comparing all 3 AI models for coding, writing, multimodal, and 6 other use cases

Jun 04, 2025

Dear subscribers,

Today, I want to share an updated guide on the best AI models by use case.

I made a video testing Claude 4, ChatGPT O3, and Gemini 2.5 head-to-head for coding, writing, deep research, multimodal and more. What I found was that the "best" model depends on what you're trying to do. Watch me test all 3 live here:

Watch now on YouTube.

Timestamps:

(00:00) ChatGPT vs. Claude vs. Gemini across 6 practical use cases
(00:29) Coding: Building Tetris in one shot
(04:01) Coding: This model built Super Mario level 1
(08:18) Writing: This model best captured my writing style
(11:15) Everyday answers: ChatGPT's memory feature is magical
(13:11) Deep research: 427 sources vs. quality insights
(18:07) Images, voice, and video: ChatGPT vs. Gemini head to head
(19:14) Live camera and computer use features you can try today

This post is brought to you by…Jira Product Discovery

PMs spend too much time updating roadmaps in static slides or collecting feedback in spreadsheets. Jira Product Discovery is here to change that, making it easy to:

Capture insights, product ideas, and feature requests in a dedicated hub.
Prioritize ideas with evidence-based frameworks.
Align on roadmaps with views for stakeholders to comment and vote.

Try Jira Product Discovery for Free

Coding: Claude is the best, but Gemini is the most cost effective

For coding, I asked each model to "Create a full-featured Tetris with beautiful graphics and controls." Here's what happened (video demo):

Claude built a gorgeous game with scores, next-piece preview, and great controls.
ChatGPT O3 created a basic clone that works but lacks features.
Gemini 2.5 made a solid game that’s not as visually polished as Claude.

To push Claude further, I asked it to "Create a 2D Mario game." After 10-15 minutes of back-and-forth, it built an actual playable Level 1 with mushrooms, goombas, and more. Neither Gemini nor O3 came close.

But here's the catch — Claude 4 Sonnet costs 20x Gemini 2.5 Flash. So if cost matters for your AI product, you might want to consider Gemini instead.

Google's been shipping like crazy lately:

Gemini 2.5 Pro and Flash are solid models for the price.
Jules is an async coding agent like OpenAI Codex.
Gemini Diffusion is a new model that lets you vibe code so fast it feels instant:

The bottom line for coding: Choose Claude 4 for the best results. Choose Gemini 2.5 for the best bang for your buck.

Writing: Claude nails your writing style

I use AI primarily to edit my writing like newsletter posts, strategy docs, and short paragraphs. My process is simple (video demo). First I give AI my best writing samples. Then I ask it to edit my draft based on those examples.

I tested all three models to edit this post using my newsletter edit prompt (available in my prompt library for paid subscribers):

Claude nailed my conversation style and format.
ChatGPT cut too much copy and lost important details.
Gemini 2.5’s edit felt too verbose and sterile.

So Claude is still my daily workhorse because it captures my writing style better than any other model — especially when I feed it examples of my best work.

Everyday answers: ChatGPT just gets you

All three models can answer everyday questions, but ChatGPT has one killer feature: Memory.

This creates magical moments. For example, it suggested I ask about the best places to visit in France because it remembered I'm planning a trip soon.

ChatGPT is also great for introspection. Try this prompt (video demo):

Tell me something unique you notice about me, but I haven't realized about myself yet. Doesn't have to be positive — just be truthful.

I'm frankly surprised Claude and Gemini still don't have memory in 2025. ChatGPT is the model that just gets you — use it to find your hidden talents and blind spots.

Deep research: ChatGPT finds the sweet spot

Every provider now offers "deep research" features, so I tested each with: "Research the AI coding market and share Bolt's strategy to compete."

The results were revealing (video demo):

Claude produced a 7-page report with 427 sources. It did a great job synthesizing insights instead of just dumping info. But the recommendations felt generic.
ChatGPT produced a 36-page report with 25 sources. It included specific recommendations that actually match what Bolt is doing — targeting non-technical users, focusing on speed, and adding integrations.
Gemini produced a 48-page report with 100 sources. It was comprehensive but the conclusions were too verbose and felt like corporate gibberish.

Here's my confession — I rarely read “Deep Research” reports in full. Instead, I skim them and then load them into an AI project to give AI the right context to polish my strategy docs and other artifacts.

For this use case, I think ChatGPT hits the sweet spot. It’s neither too short (Claude) or long (Gemini).

Image generation: ChatGPT still dominates

ChatGPT’s image feature still blows me away regularly (video).

It follows instructions the best and produces the best text rendering. I use it to create marketing assets, infographics, and even comics with my daughter.

The key is giving it examples of the style you want, then asking it for specific tweaks. Here’s a short 5 min video covering my 5 favorite use cases.

Voice chat: ChatGPT has personality (and sings)

ChatGPT has the most natural voice flow and personality. It can even sing now — terribly, which is honestly hilarious (give it a listen above).

That being said, a pet peeve I have with both ChatGPT and Gemini’s voice feature is that it likes to interrupt and cut you off. You can tell it to “Don’t interrupt me until I tell you I’m done” but I still prefer using apps like Superwhisper for voice dictation.

Video: Gemini's Veo 3 is mind-blowing

For video, Gemini is well ahead of the competition.

Google’s latest Veo 3 model turns simple text prompts into 8-second videos with sounds and voices. It’s worth a month of Gemini Pro just to experiment with. For example, I used it to make a Corgi singing at a piano.

AI video overall is poised to have a cambrian explosion in both quality and quantity. I’ll write more about this space soon.

Live camera & computer use: Not ready yet

Let's close by talking about live camera and computer use (video).

Live camera lets AI see your computer screen. I like having Gemini's live camera on while vibe coding so I can ask it to debug issues in real-time.
Computer use lets AI control your keyboard and mouse. I prefer OpenAI's Operator because it uses O3, the best reasoning model. You can use it to look up travel or make restaurant reservations.

Both features are still early, but worth trying if you're curious about AI agents.

So which model should you use?

It’s interesting how the features above reveal the each company’s strategy:

OpenAI wants to become your personal AI assistant. They're dominating the consumer market and making moves in AI coding their Codex agent and rumors of acquiring Windsurf.
Claude wants to own coding for enterprise. Their first AI conference was entirely dedicated to coding and developers. Claude is still the default model for Cursor and Anthropic is also improving Claude Code.
Gemini is finally waking up. The sheer number of AI features Google shipped at I/O is incredible. Gemini has by far the most cost-effective models and also best-in-class multimodal capabilities with Veo 3.

So here's what to use when: