Innovation Profs - 5/30/2025

Welcome to Gen AI Summer School

The neighborhood swimming pools are open, and it’s officially summer break here in Iowa. But AI education doesn’t get a summer vacation. So we’re spending the summer teaching you the essentials you need to succeed in an AI-forward world.

Here’s the plan:

• Today: Intro to large language models

• June 6: Multimedia tools

• June 13: Guide to prompting

• June 20: Building a prompt library

• June 27: Building Custom GPTs

• July 11: Intro to reasoning models

• July 18: Intro to deep research

• July 26: AI ethics

• Aug. 1: Implementing Gen AI in your job

• Aug. 8: Implementing Gen AI at your company

• Aug. 15: The road to Artificial General Intelligence

• Aug. 22: Where Gen AI is headed

Intro to Large Language Models

If you’ve been following along with our newsletters for a while, you likely know that large language models (LLMs) are the biggest driver behind the current wave of interest in generative AI.

Although natural language processing, i.e., the branch of computer science concerned with using computational tools to “recognize, understand and generate text and speech,” has been around since the 1950s, LLMs only appeared on the scene relatively recently, dating back to the 2017 paper “Attention is All You Need,” written by a team of Google researchers.

I won’t go into any of the details here about how LLMs work, but instead I’ll describe the various tools that have risen to the surface as the main players in this space. Here let’s focus on OpenAI’s ChatGPT models, Microsoft Copilot, Google Gemini, and Anthropic’s Claude models.

First, despite the Google-based origin of the idea behind LLMs, it was OpenAI that first released an LLM to the public when the model GPT-3.5 was made available via the platform ChatGPT back in November 2022. Back then, GPT-3.5 could not access the internet nor process any inputs or outputs outside of text. Since then, we’ve seen the launch of GPT-4, only accessible to paid users, which included the ability to access the web, image generation, and code generation. These two models were subsequently updated and improved and now have been replaced by faster, cheaper, and more intelligent models, including the multimodal GPT-4o (capable of handling a range of input and output types) and the lightweight GPT-4o mini. This latter model was the default model for free ChatGPT users until it was recently replaced with GPT-4.1.

For details about the various GPT models, check out OpenAI’s Models page. If you want to learn about additional ChatGPT features, consult OpenAI’s ChatGPT overview.

Next up is Microsoft Copilot. While there’s a lot more one can say about the broader features of the Microsoft 365 Copilot platform (including integration into the suite of Microsoft productivity tools), for the purposes of understanding which LLMs are available through Copilot, the answer is simple: it’s just the various OpenAI GPT models discussed above. The exact details as to precisely which GPT models you can access are unclear, as there is no interface to select which model one is using—even for paid Copilot users!—unlike the case for paid ChatGPT users. To learn more about Copilot, you can peruse the articles here.

Google Gemini is the next LLM platform we’ll consider. Although the names of both the various Google LLMs and the overall platform for accessing these LLMs have changed (anyone remember Bard?), Google has settled on Gemini for both the name of its newer models and its LLM platform (one exception is that open-source Google LLMs are the Gemma models). The currently available Gemini models are listed as version 2.5, coming in two varieties: Flash and Pro, a distinction that matches the OpenAI distinction between the mini version of a model compared to the full version of that model (e.g. GPT-4o versus GPT-4o mini). The Gemini 2.5 models also incorporate some aspects of reasoning (we will discuss reasoning models later in this series of Friday editions of the newsletter). And like Microsoft, Google has incorporated Gemini into the Google suite of productivity tools. If you want to learn more about the various Gemini models, Google’s guide on the various Google models is a good source to consult.

Finally, Anthropic is perhaps the least recognizable name in the LLM space (compared to OpenAI, Microsoft, and Google), but Anthropic is nonetheless a significant trendsetter with both their Claude family of models and advanced features such as Claude Artifacts and Computer Use. The original Claude model was only available through Slack, but Anthropic released the Claude web platform when it released the model Claude 2. Anthropic’s next move was to release three model sizes with Claude 3: Haiku, Sonnet, and Opus (ranging from small to medium to large). Claude Haiku and Claude Sonnet received updates over the previous year, but just last week, Claude Sonnet 4 and Claude Opus 4 were released, with Claude Opus 4 reportedly being the “best coding model in the world.” You can learn all about the various iterations of the Claude models here.

These four models are arguably the leaders in this space, but this is not an exhaustive list. Other LLMs worth a mention include: xAI’s Grok, Meta’s open source Llama (which can be accessed here), AI-powered “answer engine” Perplexity, French-based Mistral, and Chinese-based DeepSeek.

With all of this information, you might be asking, “Which LLM platform should I use?” For many purposes, ChatGPT is still a great option. If you want to incorporate your LLM usage into your use of office productivity tools, then Microsoft Copilot or Google Gemini are good options as well (depending on whether your organization is a Microsoft shop or a Google shop for productivity tools). Lastly, while the Claude models don’t have a large share of the market for LLM usage, they are currently the fastest growing on that front.

One last suggestion: If you really want to dig into the comparison of the various LLMs, head on over to the Chatbot Arena LLM Leaderboard to see which models are jockeying for the top spot!