💡 How to choose between GPT, Claude, & more [Free Workshop]

Hey there,

If you’re still just using one AI model for everything, you're almost certainly missing a trick.

Great results don't only come from great prompts. Often, they come from knowing which AI to prompt in the first place.

In this email, you’ll start to learn how to match the right model to each specific task by understanding the key differences between them.

In the last 6 months alone:

• Claude improved its ability to solve coding tasks by 68.4%

• GPT-4o cut response times to 0.32 seconds (nearly matching human speed)

• Gemini 1.5 started reading (and understanding) 700,000 words at once

• Llama 3 became the first free AI to match paid alternatives

At The Rundown, we’re committed to using AI to make our work easier (and better.)

Here are a few examples of how we’ve matched specific needs to the right models:

Processing Massive Documents

We all know the challenge of trying to analyze long documents with AI. You feed in a chunk of text, then another, and another... hoping the AI will understand how everything connects.

Most AI models do their best. Claude can handle about 150,000 words at once and GPT-4 processes large documents pretty effectively too. But when we wanted to test summarizing an entire book for our content, there was one standout winner.

Choosing Gemini 1.5 allows us to process:

• A full 700,000-word book in a single go

• 1 hour of detailed video content

• 11 hours of audio recordings

• Over 30,000 lines of code

Plus, even with these massive files, Gemini finds specific information with 99% accuracy.

Watch us turn an entire book into a podcast script here.

Creating Interactive Learning

Different people have different learning styles, and while language models have had a huge impact on the way we can consume information, there are some limits to the traditional chatbot experience.

While every modern AI can help create educational content, most just generate more static material. Gemini creates comprehensive outlines, GPT-4o writes detailed explanations, and Llama helps structure the content. But we needed something that could make learning feel more interactive.

Using Claude's Artifacts feature, we can:

• Transform any workshop into a quiz

• Generate questions that test real understanding of the material

• Include detailed explanations that cite the source content

• Publish and share instantly with one click

Best of all, it maintains impressive accuracy even with complex technical projects.

See how we turned a popular workshop into an engaging game here.

Building Mobile AI Assistants

Remember when having an personal assistant on your phone meant being locked into someone else's system? Traditional options aren’t exactly ideal when you want an AI assistant that thinks and responds your way.

The usual suspects all offer some customization. Gemini lets you adjust parameters, Claude gives you control over responses, and GPT-4 has its API. But we wanted something that could run right on our phone and respond instantly to any request.

With Llama 3, we built an assistant that:

• Runs directly on a device (like Siri)

• Responds instantly to voice commands using a lightning-fast API

• Can be completely customized to your needs

• Handles everything from coding challenges to trivia questions

And because it's open source, you can modify it to work exactly how you want.

Watch how we created an on-device Siri upgrade that actually understands what we're asking for here.

Delegating with Training Systems 

If you’ve ever tried to help people in your team learn a particular process, you’ll know it usually goes something like this: Record a video, write some instructions, answer the same questions over and over... and still run the risk of them missing crucial details.

Of course, Gemini is great at breaking down complex topics, Claude creates well-written explanations, and Llama is capable of brilliant responses. However, there was a much better option for turning team knowledge into interactive training.

With GPT-4o, it’s possible to:

• Turn any Loom video transcript into a training guide

• Build a CustomGPT on the exact process from the video transcript

• Get team members to guide themselves through the right steps

• Build a solution that can adapt to different learning styles

And because it has access to our information, it can answer questions exactly how we would.

Watch us turn a 7-minute video into a complete training system here.

As these examples show, there's no "best" AI model. 

But there is the right tool for each job. 

When making a choice, you need to think about things like document length, processing speed, costs, rate limits, accuracy, writing skill, integration options, infrastructure requirements, multimodal capabilities, reasoning abilities, fine-tuning options, data privacy concern, and ecosystem support. 

And if that all sounds exhausting, you’re in luck…

This Friday at 4 PM EST, meet Dr. Alvaro Cintas to learn:

• The key factors to consider when choosing between models

• How to evaluate model performance for specific tasks, from content generation to data analysis.

• Real-world examples and a comparison guide to help you make an informed decision.

Building with the wrong model can limit your outcomes, but by the end of this workshop you’ll be crystal clear on how to make the right choice every time.

Plus, showing up live also gives you the opportunity to ask about your own use case in the chat and get important insights that you can implement straight away.

Grab a free 14-day trial for The Rundown AI University to make the most of the different options.

See you in class,

Rowan & The Rundown Team

P.S. The AI landscape moves fast. See how top performers are choosing (and using) their models by unlocking a 28-day trial to The Rundown AI University here.