AI Advances Alarm Attorneys: Opas 4.6 Raises Legal Benchmark
Just a month ago, the prevailing notion was that the legal profession was still safe from the immediate reach of artificial intelligence (AI). Professional tasks—especially legal and corporate analysis—had long been considered beyond AI agents’ capabilities, as large research laboratories’ scores on Merrkor’s new benchmark remained below 25 percent. However, in just a few weeks, that scenario has dramatically shifted, according to TechCrunch.
This week, with the release of Anthropic’s new model Opas 4.6, a significant leap has been observed in AI performance scores. In a single evaluation, the model scored nearly 30 percent, while the average score across multiple attempts rose to 45 percent.
Analysts attribute this jump to the ‘agentic’ features introduced in the new version—particularly the ‘agent swarm’—which aids in solving complex, multi-step problems.
Brendon Foody, CEO of Merrkor, described the progress as “incredible”, noting that moving the score from 18.4 percent to nearly 30 percent in just a few months is astonishing.
While the road to 100 percent capability remains distant, this advancement demonstrates that the pace of improvement in foundational AI models is far from slowing down. The implication is clear: while lawyers are not immediately at risk of losing their jobs, the previous sense of security in the profession is no longer guaranteed.
DBTech/BMT/OR
