Vibe Check: Claude Sonnet 4.5

Was this newsletter forwarded to you? Sign up to get it in your inbox.

Anthropic just rolled out Claude Sonnet 4.5, and, of course, we spent the weekend using it to code and running long agentic tasks with it.

The headline: It’s noticeably faster, more steerable, and more reliable than Opus 4.1—especially inside Claude Code. In head-to-head tests it blitzed through a large pull request review in minutes, handled multi-file reasoning without wandering, and stayed terse when we asked it to.

It won’t dethrone GPT-5 Codex for the trickiest production bug hunts, but as a day-to-day builder’s tool, it feels like an exciting jump. Here’s our day zero vibe check.

Speed

If you’re used to using Opus in Claude Code or the Claude app, you’ll be happy: The new Sonnet 4.5 is really fast. Kieran Klaassen, general manager of Cora, said, “It feels about 50 percent faster than previous versions of Claude.”

In a head-to-head code review challenge, it finished a comprehensive code review of a new feature in a large code base in about two minutes. GPT-5 Codex took about 10 to do the same task.

Speed is a dimension of intelligence, and Sonnet 4.5’s speed makes it much easier to pair with.

Performance

It’s quite good at long-running agentic tasks in the Claude app and in Claude Code. I fed it the three spreadsheets we use to run Every, our profit-and-loss accounting, our weekly performance tracker, and our consulting tracker—and it easily wrote a Word doc with a third-quarter investor update that I could’ve sent with only minor tweaks.

Make your team AI‑native

Scattered tools slow teams down. Every Teams gives your whole organization full access to Every and our AI apps—Sparkle to organize files, Spiral to write well, Cora to manage email, and Monologue for smart dictation—plus our daily newsletter, subscriber‑only livestreams, Discord, and course discounts. One subscription to keep your company at the AI frontier. Trusted by 200+ AI-native companies—including The Browser Company, Portola, and Stainless.

Create your team

Want to sponsor Every? Click here.

Kieran found that it solved a bug in Cora in about 20 minutes that Opus 4.1 couldn’t crack at all. He also used it to vibe code an iOS app for Cora by feeding it the current codebase and a book on iOS programming:

This jump in performance seems to be a combination of:

Become a paid subscriber to Every to unlock this piece and learn:

1. The four ways in which Sonnet 4.5 outperforms previous models
2. The results of the Reach Test from Dan, Kieran, and Every's Alex Duffy
3. The final verdict: Sonnet 4.5 versus Opus 4.1 versus GPT-5