Artificial intelligence agents, once deemed far from professional competence, have made a significant leap forward. Anthropic's recently released Opus 4.6 model has dramatically reshaped performance benchmarks set by Mercor, particularly in complex professional tasks such as legal and corporate analysis. This rapid advancement suggests that the timeline for AI displacement in professions previously thought secure might be accelerating.
Just last month, a new benchmark by Mercor measured AI agents' capabilities across various professional tasks. At the time, scores were notably low, with major AI labs scoring under 25%. This led to a consensus that lawyers, for instance, were safe from immediate AI displacement.
However, the pace of AI development can be astonishingly fast. This week's release of Anthropic's Opus 4.6 has significantly shaken up the AI agent leaderboards. Anthropic's new model achieved nearly 30% in one-shot trials and an impressive 45% when given multiple attempts at solving problems. A key factor in this improvement appears to be the inclusion of new agentic features, such as "agent swarms," which are designed to enhance multi-step problem-solving.
This score represents a substantial jump from the previous state-of-the-art, signaling that progress in foundation models is not slowing down. Brendan Foody, CEO of Mercor, expressed his astonishment, stating,
"jumping from 18.4% to 29.8% in a few months is insane."
While a 30% or 45% score is still a considerable distance from 100%, implying that lawyers won't be replaced by machines next week, the rapid progress is undeniable. Professionals across various fields, particularly those in law, should now be significantly less confident in their immunity to AI displacement than they were just a month ago.
