Opus 4.8 is out across Claude products at the same price as before. Key additions include a fast mode running at 2.5x speed and 3x lower cost, dynamic workflows supporting hundreds of parallel subagents in Claude Code, and effort control in the Claude app.
Opus 4.8 is available in Claude Code today, with a thread of key details covering benchmarks, effort modes, rate limits, and long-horizon task guidance.
Opus 4.8 scores 69.2% on SWE-bench Pro, up from 64.3% on Opus 4.7. It is also around four times less likely than 4.7 to let flaws in its own code pass without flagging them.
Opus 4.8 defaults to high effort but uses similar tokens to the 4.7 default for typical coding tasks while delivering better results. For difficult or long-running async work, an xhigh effort mode is available, and rate limits in Claude Code have been raised to match the increased token usage.
Dynamic workflows (research preview) let Claude write an orchestration script on the fly and spin up a large fleet of coordinated subagents in parallel to tackle complex tasks. Say "workflow" in a prompt to trigger it.
Mentioning "workflow" in a prompt triggers Claude to generate a structured orchestration plan it then strictly follows, ensuring each stage completes in the correct order. The feature is framed as the most significant new addition to Claude Code.
Dynamic workflows are available today across Max, Team, Enterprise, and the API (including Bedrock, Vertex AI, and Foundry). On by default for Max and Team plans; Enterprise admins enable it via managed settings.
Dynamic workflows are designed for work that outgrows a single agent loop: think service-wide bug hunts, large migrations, or stress-testing a system design. They can burn through tokens quickly, so starting with a scoped task is strongly recommended.
Dynamic workflows are reusable: save one as a slash command inside a project to share it with teammates, or in your home directory to make it available everywhere.
A blog post covers how dynamic workflow orchestration works under the hood and showcases what early users have already built with it.
The standout new capability in this release is dynamic workflows: Claude now plans a task, fans it out to tens or hundreds of parallel subagents, verifies their outputs, and iterates until everything converges into a single coordinated answer.
Big refactors and migrations, the work teams perpetually defer, can now land in days or weeks instead of tying up engineers for a quarter. The announcement links to more detail on dynamic workflows.
A concrete internal use case: dynamic workflows were used to catalogue hundreds of A/B test flags and identify ones rolled out to 0% or 100% for deprecation, with subagents working in parallel rather than sequentially.
A new effort level called "ultracode" runs at xhigh and lets Claude autonomously decide when a task is complex enough to warrant spinning up a dynamic workflow, without you having to prompt for it explicitly.
Dynamic workflows combined with adversarial code review were the key techniques behind rewriting Bun in Rust in six days, offered as a concrete proof point for the approach.
Jarred Sumner used dynamic workflows to port roughly 750k lines of Bun from Zig to Rust in 11 days, reaching 99.8% test suite pass rate, with hundreds of agents running in parallel and two reviewers assigned to every file.
A reliable three-phase loop for each unit of work: do the work (no slow commands), run adversarial review, then apply changes. Works for bulk rewrites like porting every .zig file to Rust or getting a whole crate tree to compile.
A concrete workflow pattern: step one fixes a bug from a report file, step two spins up two adversarial review agents to poke holes in the fix. Git and build commands are banned per step to avoid conflicts when multiple Claude instances share a branch.
You can tell Claude roughly how many tokens to spend, nudging it toward thoroughness on hard tasks and frugality on easy ones. This works especially well inside dynamic workflows.
