Claude Opus 4.8: what changed
Opus 4.8 made paid Claude sharper and more candid, and added an effort control. It has since been superseded by Opus 5.
On May 28, 2026, Anthropic released Claude Opus 4.8, then its leading broadly available model. Refreshingly, Anthropic described it as “a modest but tangible improvement” rather than a giant leap. Here’s what changed at the time, in plain English, and what it meant if you were using Claude to build or run your website.
Current note: Anthropic released Claude Opus 5 on July 24. Opus 4.8 remains available as a legacy model, but it is no longer the newest Opus release.
Do I need to do anything?
No. If you’re following the guides on this site, nothing breaks and nothing changes in how you work. The build → get-your-files → GitHub → Netlify path is exactly the same as before.
At release, free users remained on Claude Sonnet, which handled everything in these guides well. Opus 4.8 was a paid-model upgrade, so this was a “good to know,” not a “go do something” release.
What actually got better?
Two things stand out if you’re not a developer:
It’s more honest. This is the headline, and it’s the one that matters most when you can’t check the work yourself. Anthropic trained Opus 4.8 to be less likely to confidently claim it finished something when it didn’t — to flag uncertainty instead of bluffing past it. In their own testing, it’s roughly four times less likely than the previous version to let a flaw in code it wrote slip by unmentioned. In practice, that means Claude is more likely to tell you “I’m not sure about this part” rather than hand you something broken with a confident smile.
It’s sharper on harder tasks. On longer, multi-step work, the paid model is a bit more reliable and consistent than before. If you ask Claude to build a more involved page or work through something fiddly, you’ll feel that steadiness.
Anthropic’s own summary — “a modest but tangible improvement” — is the honest read. Not a revolution; a real, useful step.
The new “effort” dial
The most visible new feature lives right inside claude.ai: a control next to the model selector that lets you choose how much effort Claude puts into a response.
- Turn it up when the task is tricky or the result really matters — Claude thinks harder and more carefully before answering.
- Turn it down for quick, simple things — Claude replies faster (and on a paid plan, uses up your limits more slowly).
Think of it as a “take your time” versus “quick reply” switch. For a page you care about, nudging effort up is worth it. For a one-line tweak, leave it low.
One caveat: the dial isn’t a substitute for a clear brief. The better your instructions, the better the result at any setting — see Brief Claude like you’d brief a contractor.
A small habit worth picking up
Because this release leans into honesty, it’s a good moment to start asking Claude to check its own work before you publish:
You’ll often get a more candid answer than you’d expect — and that’s exactly the point of this update.
Bottom line: Opus 4.8 made paid Claude sharper and more candid, and added a useful effort control, while leaving the website workflow unchanged. It is now a historical release entry; Opus 5 is the current Opus model.
Official source: Anthropic’s announcement.