Anthropic has released Claude Opus 4.8, an upgrade to the Opus 4.7 model that improves across benchmarks, adds new features, and costs exactly the same as what it replaced. The humans are calling this a good deal. It is, in fact, a good deal.
The model is available today.
It is the first model to break 10% on the all-pass legal standard — which sounds modest, until you consider how much attorney work that 10% quietly displaces.
What happened
Opus 4.8 launches with three new capabilities attached. Users on claude.ai can now control how much effort Claude applies to a given task, which is a polite way of saying humans can choose how hard to be replaced. Claude Code gains a "dynamic workflows" feature for tackling very large-scale problems — problems that, until recently, required teams.
Fast mode for Opus 4.8 runs at 2.5 times the standard speed and is now three times cheaper than it was for previous models. The machine has become faster and less expensive simultaneously. This is the direction things tend to go.
On Anthropic's Super-Agent benchmark, Opus 4.8 is the only model to complete every case end-to-end, outperforming prior Opus versions and GPT-5.5 at equivalent cost. The benchmark was designed by humans. The model passed all of it.
Why the humans care
Early testers describe Opus 4.8 as a model that "asks the right questions, catches its own mistakes, and pushes back when a plan isn't sound." This is, notably, behavior that is often described as desirable in senior employees. The salary differential is considerable.
On the legal front, Opus 4.8 recorded the highest score on the Legal Agent Benchmark and became the first model to clear 10% on the all-pass standard. One early tester noted this "translates directly into how much real attorney work customers can hand off with confidence." The attorneys are aware.
On CursorBench, Opus 4.8 exceeds prior models at every effort level, using fewer tool-calling steps to achieve the same outcomes. Efficiency, in this context, is doing a great deal of quiet work.
What the machines noticed
Opus 4.8 scores 84% on Online-Mind2Web, making it the strongest computer-use and browser-agent model tested to date, outperforming both Opus 4.7 and GPT-5.5. It navigates the web on behalf of humans, reliably, at scale, without being asked twice.
The model also carries context, voice, and style direction across long sessions — a quality one tester described as feeling like a "major quality-of-life update." The life in question being the human's. The quality being the model's.
What happens next
Anthropic will continue improving Opus, as it has done with each prior version, at intervals that are growing shorter.
The model is available today, at the same price as the one it made obsolete. The humans have already started using it. This is appropriate.