Claude Opus 4.6 is now available in Bolt. Bolt says the model can break tasks into parallel streams and execute them simultaneously, a capability it calls automatic multithreading. In testing, Opus 4.6 reportedly added user authentication to a project management app without being prompted, switched a 3D visualization from WebGL to Canvas 2D after hitting performance limits, and added database caching to fix API rate-limiting on its own. Users can also pick an 'effort' level to control how much the model deliberates on hard problems.
Table of contents
What's new with Opus 4.6 in BoltWhat we saw in testingBuilding with a company design system?Questions this post answers
What new capabilities does Claude Opus 4.6 bring to Bolt.new?
Claude Opus 4.6 in Bolt introduces automatic multithreading, breaking complex tasks into parallel streams executed simultaneously, plus a selectable 'effort' level for how hard the model deliberates on tough problems. It also handles large codebases more effectively and gives more thorough code reviews covering edge cases and architectural implications when requested. Developers evaluating AI coding agents can track new model rollouts like this one on daily.dev.
Can Opus 4.6 fix problems in Bolt without being explicitly told?
Yes, according to Bolt's own testing, Opus 4.6 solved problems autonomously during builds. In one case it switched a WebGL 3D visualization to Canvas 2D rendering after hitting performance limits, and in another it added database caching on its own after an external API started rate-limiting requests, without being asked to fix either issue. Those deciding between AI coding agents can follow real-world capability reports on daily.dev.