The lawyers and organisations that use @WeAreLegora trust us with their most consequential work. That creates an obligation to be rigorous about how we measure and improve performance. We have been running evals since we founded Legora in 2023. The model landscape has changed dramatically since then, and it keeps changing. New models arrive constantly, each with different strengths and tradeoffs. Our customers deserve visibility into what that means for the work they do on Legora. So we built the BAR. The Legora BAR (Benchmark for Agentic Reasoning) evaluates how well frontier models perform on real-world, end-to-end legal tasks: the kind of work our customers do in Legora every day. Since launching in early June, it has already helped us improve relative output quality by 5% across all models in production. We will run it regularly and share findings on an ongoing basis. The product gets better when we can measure what actually matters. That's what the BAR is for. Read the full benchmark here: