Ultra Mode Preview — Pushing Recall to the Next Level

[lizhengfeng101]

On the AACR-Bench benchmark, Open Code Review already leads in F1 score, precision, time efficiency, and token cost.

We honestly acknowledge, however, that our recall is not yet leading — and we want to close that gap.

We are developing Ultra Mode, a deeper review pass that trades additional time and token budget for significantly broader issue coverage. Our goal is to meaningfully improve recall while preserving the precision and efficiency advantages that define Open Code Review.

This feature requires extensive benchmarking and validation before release. We will share progress updates as the work advances.

Stay tuned.