← All writing

NewsletterAI

Google's new model is worth testing

Last month I told you not to trust Google's marketing. This time, the model holds up.

Google just launched Gemini 4 Neon. Like every Google launch, at first glance it looks like the best model in the world.

Usually that falls apart once people use it on real work. It did last month with Argon, Google's previous model.

This time, it might actually be true.

I usually ignore Google's models completely. But people I trust tried Neon and gave it rave reviews. They told me it was nearly as good as Claude, and it costs less. I was skeptical, so I tried it myself. I have to say, they were right.

Plenty of people still prefer Claude for long jobs, so test it on your own work. It takes about ten minutes.

Here's what I'd do if I were you. Pick three tasks this week that the AI you use every day nails, and try them on Neon. Maybe you'll like it better on all of them, or maybe you'll like it on some and not others. Either way, you'll learn a lot, and you might end up with a new model to use every day.

You can get Neon in the Gemini app on the Pro and Ultra plans, and in Gemini for business.

If you're using it for engineering work, make one of the three a hard task, the kind your current AI barely gets right.

Hit reply and tell me which model won.

— Matt

What should I write about next? Vote or suggest a topic →

Get the next essay in your inbox.

Reviews, guides, and the roundup. Free. No filler. The next one lands the day it's ready.