From the source
Legora published the October results of its Benchmark for Agentic Reasoning, which tests frontier models on tasks drawn from real-world legal cases.
- The update adds filters for practice area, task family and model, letting readers explore performance across 30 practice areas, seven task families and 17 models.
- Legora says Opus 5.5 leads on overall quality this month at 6% above average.
- Overall rankings do not capture everything: Muse Spark 1.3 ranks 10th overall but places second on review tasks, which Legora cites as a reason to filter by task family.
- Legora says it has continued testing the latest frontier models since launching the benchmark, tracking what improves legal work as models and its harness evolve.
