Dear subscribers,
Today, I want to share an opinionated guide on which AI model you should use.
The AI landscape has changed dramatically since I shared my tier list of the top 24 AI tools last year. Multiple capable AI models are now competing for your attention.
So which model deserves your time and money? Let’s break it down below.
This post is brought to you by…Vanta
Join over 9,000 companies like Atlassian, Factory, and Chili Piper who streamline compliance with Vanta’s automation and trusted network of security experts.
Get $1,000 off Vanta by booking a demo via my exclusive link below.
ChatGPT: The swiss army knife of AI models
If you want the best multimodal AI assistant, ChatGPT is your best bet.
ChatGPT is best for…
ChatGPT is great for random tasks like comparing window quotes
- Answers to everyday questions. I ask ChatGPT many questions every day, and it never hits a rate limit. For example, I’m replacing the windows at my house and asked ChatGPT to “Compare these two PDF quotes and tell me which deal is better.” It gave me an incredible breakdown.
- Voice conversations. I love pasting some context into ChatGPT and then activating voice mode to talk to it during a walk or commute.
Pro tip: If you don’t want voice mode to interrupt you, say: “Please don’t respond until I say I’m done.”
- Live camera with vision and voice. While in voice mode, you can activate your phone camera to let ChatGPT see what you’re seeing. I’ve found it incredible to learn new things with my kids.
- Math and financial analysis. I’ve asked ChatGPT to help with tax prep, RSU planning, and more. It seems to be better at math than other models.
- Deep research on a market or technical topic. I’ve written about how Deep Research is ChatGPT’s best product. It outperforms similar offerings from Grok, Perplexity, and Gemini—and now it's available for $20/month.
ChatGPT is not the best for…
Editing writing or writing code. I’ve found ChatGPT’s writing suggestions to be very mechanical (although GPT 4.5 is better). It’s also not the go-to model for AI coding platforms. That honor still belongs to…
Claude: The writer and coder’s preferred choice
Claude doesn't have all the bells and whistles of ChatGPT, but it does writing and coding (80% of my AI usage) incredibly well.
Claude is best for…
- Editing your writing. Give Claude your best writing samples, and it will quickly adapt to your style. Its ability to create a first draft from my unpolished voice notes (via Superwhisper) is a massive time saver.
Claude works amazingly well with my “edit newsletter” prompt
Enjoying this article? Unlock my edit newsletter prompt and the 20+ other prompts in my AI prompt library by upgrading to paid.
- Vibe coding. There’s a reason why AI coding platforms like Bolt and Cursor use Claude 3.7 Sonnet as their default model. I’ve been “vibe coding” games (e.g., a plane simulator, Star Wars Death Star run, and a zombie first-person shooter) with it, and it’s been incredibly fun.
Claude is not the best for…
Multimodal capabilities. Claude lacks image, voice, and video generation. However, its new extended thinking mode is comparable to OpenAI's best reasoning models.
Gemini: Great if you need long context and video
Veo is the best at AI video and I love using it to make short movies with my kids
Gemini offers the longest context window of all AI models and is amazing for video summaries and generation.
Gemini is best for...
- Making videos with Veo. Veo is miles ahead of OpenAI’s Sora model. My daughter and I love using it to generate short movies like this story about a girl and a hamster who lose each other on the Golden Gate Bridge.
- Creating images with Imagen 3. While DALL-E and Midjourney get more attention, Gemini's Imagen 3 often produces more coherent and artistic results. I love how it handles complex scenes and maintains consistency.
- Processing massive documents and long video uploads. With a 2M+ word context window, Gemini can simultaneously process entire books, multiple documents, and even video uploads.
Gemini is the only AI model on the list that processes video uploads
Gemini is not the best for…
Personality. This is a weird complaint, but Gemini’s responses are often too brief and clinical (even more so than ChatGPT). I’ve asked it to edit my blog post and write code, and it just doesn’t feel as great to use as Claude or ChatGPT.
Grok: The best free (and unhinged) AI model
Grok is the best free AI model available if you don’t want to pay $20 / month. It also has the most up-to-date information along with Perplexity.
Grok is best for...
- Free deep research, reasoning, image generation, and coding. Grok offers free versions of Deep Research (“DeepSearch”), reasoning (“think”), image generation, and coding that are best-in-class for a free AI model. I think Grok’s coding is comparable to Sonnet 3.7 but lacks integration with AI codegen tools.
- Finding information on X (Twitter). Grok is the only model that has access to real-time posts from X (Twitter).
- Unhinged voice conversations. Grok’s voice mode is a thing to behold. Try its sexy or unhinged voice modes away from your family. I don’t know why they shipped these voices, but they’re quite something.
Grok is not the best for…
Editing writing. Grok is terrible at editing your writing because it has a default sarcastic and witty style that’s frankly inappropriate for any real writing task.
Perplexity: Quick research, trip planning, and more
Perplexity gave me personalized restaurant recommendations for my Japan trip
Perplexity doesn’t pre-train its own AI models, but I still included it here because the Pro version also costs $20 / month, and it’s my Google Search replacement.
Perplexity is best for…
- Quick research. Need to understand a topic quickly? Perplexity pulls information from the web and presents it in clear, concise summaries with direct source links. I use it to look up facts and quick overviews.
- Travel planning and local activities. Perplexity is surprisingly good at crafting detailed travel itineraries and finding local events. Recently, it helped me plan my upcoming trip to Japan.
Perplexity is not the best for…
Deep research. I’ve found that ChatGPT and Grok’s Deep Research look at more comprehensive sources than Perplexity’s.
Bonus: DeepSeek is best for getting a good laugh
Remember all the hype around DeepSeek from China? I still love looking at its raw chain of thought to get a laugh.
For example, how it tries to reason through my dad jokes cracks me up:
Try it when you want to have fun with your family or team.
So, which model should you spend your $20 on?
Here’s how I’m thinking about it:
- If I had only $20, I’d spend it on Claude. Editing writing and (to a lesser extent) coding are the use cases that account for 80% of my AI usage. Claude is the daily workhorse that helps me get work done.
- If I had another $20, I’d spend it on ChatGPT. I love its voice mode, reasoning, and deep research capabilities.
Let me know in the replies what you think and if you agree with my choices :)
