The most powerful Anthropic Claude model is growing, the company says in a blog post Thursday that Claude Opus 4.6 will be even better at coding and creating projects on the first go.
Claude Opus 4.5 is already a strong model for coding, with its release in November evoking Claude Code’s viral vibe-coding moment over the holidays. Claude’s proven coding abilities and the new Cowork feature have Wall Street worried, as many tech stocks have fallen in recent weeks, due to concerns that people won’t need software products in the future.
Anthropic said the new model is more focused on solving big challenges, such as the inner workings of complex applications, while also handling simple steps quickly.
As a conceptual model, Opus 4.6 works by breaking down the steps it needs to take to do what you’re asking and putting the program together before starting. It will also go back and check its work on those steps, sometimes making multiple attempts without you asking.
Sometimes the model can use too much effort on a task, which Anthropic said can be solved by reducing its effort level from the default “high” setting.
Read more: Anthropic Super Bowl Commercials Pinky’s Promise No Ads to Claude
Claude Opus models are available for paying Claude users in the Pro, Max, Team and Enterprise plans. The cheapest of those, Pro, costs $20 per month (or $17 per month if you pay annually). The Pro edition comes with restrictions on the use of Opus, which users can hit after a few hours of coding vibe and then have to wait a few hours for it to reset.
Other than Opus, Anthropic has smaller models, with less power to them Sonnet 4.5 again Haiku 4.5.
First look at Claude Opus 4.6
To test the new model, I tasked it with creating a voice-activated trivia app. This process took several iterations over the course of an hour, but Claude pulled each one off very quickly. It wasn’t independent — I pointed out errors and suggested solutions, although some of my suggestions failed because we were up against the constraints of building within an HTML file entirely.
Enlarge Image
The app Claude Opus 4.6 built for me really depended on Jeopardy! question style.
The experience wasn’t much different when I tried the same test with Opus 4.5, although this seemed to run a little faster. The model captured the idea of what I was trying to do from the start, which isn’t always the case with AI projects, and the trivia questions it came with, once I told it to trick it, were well designed. Most of them were correct, too, although one of the (many) art history questions asked me to name the artist (Edvard Munch) but he told me that the correct answer was the title of the painting (The Scream).
The downside of the speed is that I burned through the usage limit on my Pro system for about 90 minutes — just as I got the app running seamlessly — and I couldn’t do one last request: for a database of over 100 questions. That will have to wait another few hours.
