Opus 4.5 yields the best precision of any model tested for code review (blog.macroscope.com)

Macroscope has launched an upgraded version of its Code Review tool, now powered by Claude Opus 4.5, which delivers unparalleled precision for identifying software bugs. According to internal benchmarks, this new model outperforms its predecessor, GPT-5.1, with a remarkable 40% increase in recall and 10% reduction in false positives, significantly enhancing its ability to detect valid issues. With a 25% higher F1 score, a metric combining precision and recall, the new pipeline represents a significant step forward in automated code reviews. The significance of Opus 4.5 lies in its ability to balance performance with cost and latency, delivering superior accuracy in bug detection while still maintaining user-friendly response times for CI/CD processes. Although the latency is about 40% higher than the previous version, averaging 262 seconds, users have reported that the benefits of improved recall and precision justify this trade-off. This update not only promises to improve software quality by minimizing false alerts but also sets a new benchmark in the AI-driven code review landscape, encouraging developers to consider Macroscope for their code quality needs.