That was certainly the case with Cloud Ops 4.5, the latest version of Anthropic’s most powerful model, released in late November. In December, Mater announced that Opus 4.5 appeared capable of independently completing a task that would have taken a human about five hours. A human security researcher tweeted that he would change the direction of his research in light of the findings. Another employee of the company simply wrote, “Mom come pick me up I’m scared.”
But the truth is more complicated than these dramatic reactions. For one thing, estimates of the capabilities of specific meter models come with considerable error bars. As Mater clearly stated on X, Ops 4.5 might only be able to complete tasks in two hours, or it might succeed in tasks that would take humans up to 20 hours. Given the uncertainty surrounding this method, it was impossible to know for sure.
“There are a bunch of ways that people are reading too much into graphs,” says Sidney Van Arx, a member of MetRK’s technical staff.
More fundamentally, MetRplot does not measure AI capabilities, nor does it claim to. To create graphs, MetR tests models primarily on coding tasks, measuring the difficulty of each one or estimating how long it takes humans to complete. A metric that not everyone accepts. Cloud Ops 4.5 may be able to complete some tasks that would take humans five hours, but that doesn’t mean it’s anywhere close to replacing a human worker.
MetR was founded to assess the threats posed by frontier AI systems. Although it is best known for its austerity plot, it has also worked with AI companies to evaluate its systems in more detail and has published several other independent research projects, including a The July 2025 study covered a wide range suggest that AI coding assistants are actually undermining software engineers.
But the parsimony plot has built the meter’s reputation, and the organization seems to have a complicated relationship with the graph’s often breathless reception. In January, Thomas Cova, one of the lead authors of the paper, who introduced it, Wrote a blog post Responding to some of the criticisms and clarifying its limitations, MetR is currently working on a more extensive FAQ document. But KWA doesn’t expect these efforts to replace meaningful conversations. “I think the hype machine will basically, whatever we do, just remove all the caveats,” he says.
However, the MetR team thinks that there is something meaningful about the speed of the AI’s progress in the plot. “You shouldn’t tie your life to this graph at all,” Van Arx says. “But also,” he added, “I bet the trend continues.” “