Interested in more of Jen’s thoughts on AI? Catch episode 24 of the What’s Up, Wake podcast: Politely Breaking Down AI with Jen McFarland.
Fall doesn’t just signal the comeback of Pumpkin Spice Lattes. This November will mark three years since the release of OpenAI’s ChatGPT, which reached 1 million users in five days and launched a thousand new AI models.
In these three years, AI has been all over the news, and the pace of continued development has been frantic. AI usage has been increasing, with the market growing by about 30% per year, and a rate of adoption higher than mobile phones in their early years. However much of that growth is represented by younger generations, and some consumers have been slow to pick up generative AI, particularly outside of work.
Let’s take this moment to look at the current state of things, and where you can start if you’re feeling AI-curious.
Comparing the Options
While new models are being developed and improved all the time, three companies have emerged above the rest, offering quality models, evolving on a regular basis, and consistently performing well. These are Anthropic, Google, and Open AI.
For the most part, you won’t go wrong using any of these three services, which are in a statistical dead heat for favorite according to the LMArena AI Leaderboard for user favorites.
gemini.google.com
Latest models: 2.5 Pro, 2.5 Flash, Nano Banana (image generation)
Hallucination of top models: 1.1% *
Cost: $20/month (See all pricing)
Environmental impact data **
Open AI
chatgpt.com
Latest models: GPT-5 High, GPT-5 Chat, Sora (image and video generation)
Hallucination of top models: 1.25%
Cost: $20/month Personal; $25/ month Business (See all pricing)
Anthropic
https://claude.ai/
Latest Models: Clause Opus 4.1, Claude Sonnet
Hallucinations average of top models: 4.8%
$17/month (annually) https://claude.ai/upgrade
* https://huggingface.co/spaces/vectara/leaderboard
** To date, only Google has officially released environmental impact data on their models. While estimates have been made for Anthropic and Open AI models, the companies have not shared this data.
Choosing a Model
One thing that should be a concern when choosing a model is your use case. Before you start any prompting, consider what you’re trying to do and see if you can pick a model based on your needs. Are you doing something simple and text based? Go with a “quick” or basic model. Do you need to work with media or formats other than text? Make sure the model allows uploading of resources. Are you using AI as a Google search replacement or doing topic research? Check out Perplexity, or consider other application-based AI tools. Examine all of your needs before you get started to save yourself time and reduce your impact.
That said, selecting the right AI tool can be a challenge thanks to model naming conventions — or lack thereof. OpenAI in particular has confused many of us with model names like “o3” and “4o” (where ironically 4o is less capable than o3), and qualifiers including “low,” “high,” “pro,” and “minimal.”
To help address this issue, OpenAI’s most recent offering, ChatGPT-5, is intended to route users to the best model for them based on their prompt. So if you have a short, text-only question, your response will come from a simpler and faster model; if you submit a math or research query, your response may take longer and be completed by a more advanced model. Google now has a similar solution with a toggle between two logically named options: “Flash” and “Reasoning.”
All of these options have free versions, and it’s a good idea to test out multiple models before you pick a favorite. Remember, however, that when using a free model your interactions are almost certainly being used to train the model. In many cases, the content you submit will be saved by the company for potential future use. Always read and review an app’s privacy policy before engaging.
If you want to use models without privacy concerns, you’ll need to pay for the service, but again — make sure you read the rules and confirm that your settings are properly configured to keep your content private.
Wrapping Up
Remember that this is information for right now, and AI will continue to evolve and improve. Many of the links here are updated as new models are released, so keep an eye on the data! Over time you’ll get a sense for working with the different models, and you’ll probably develop a preference based on your interactions.
