The best AI model ever?

Good morning!

Today, all I want to talk about is Claude 4. It dropped, and broke the internet (literally, but briefly).

But before we begin, a few words from our sponsor:

Start learning AI in 2025

Keeping up with AI is hard – we get it!

That’s why over 1M professionals read Superhuman AI to stay ahead.

• Get daily AI news, tools, and tutorials

• Learn new AI skills you can use at work in 3 mins a day

• Become 10X more productive

Sign up and start mastering AI

Here’s what happened:

The Main Event: Claude 4 Is Here and It's... Actually Really Good?

What happened: Anthropic released Claude Opus 4 and Sonnet 4, and the hype train left the station at light speed.

We're talking 72.5% on SWE-bench (the coding benchmark that makes engineers cry), extended thinking that can run for hours, and tool use that actually works.

Why it matters: This isn't just another "we made the numbers go up" release.

Claude 4 can maintain context across marathon sessions, work with tools during its thinking process, and – get this – it follows instructions so well it's almost annoying. Like that overachiever in your group project who actually reads the rubric.

Here's where it gets spicy. 

Alex Albert, dev relations guy from Anthropic (so take it with a pinch of salt) had something to share on Twitter.

According to him, they had to reduce their web search prompt by 74%. Why? Because Sonnet 4 was too good at following instructions.

The old model would ignore half their messy examples, but Sonnet 4 faithfully reproduced every inconsistency in their prompts.

Peak 2025 energy right there.

Market Reality Check: Still Team Google?

But Prediction markets don’t think Claude is the real winner for May.

Prediction markets briefly spiked Anthropic's chances from 2% to 31.6% within minutes of the release. But now?

We're back to betting on Google having the best overall model.

This feels like the classic "new toy syndrome."

Claude 4 is genuinely impressive, especially for coding and long-form reasoning. But Google's sitting on some serious compute and hasn't played all their cards yet.

The real winner? Us.

Competition is heating up, and these models are getting scary good at actually being useful instead of just impressive in benchmarks.

Bottom line: Claude 4 is the real deal, but don't count out the other players yet.

The AI race is far from over, and that's exactly how we want it.

P.S. If you're a developer, Claude Code is now generally available. Time to let AI pair program with you while you grab coffee. What a time to be alive.

🤝 Share Get Into AI with your friends!