Release details

New

• Native Benchling ELN integration — sync candidates, results, and protocols directly from your lab notebook

• Rebuilt explainability interface — SHAP attribution now displayed inline with each prediction, not in a separate export

• Batch screening API endpoint — run up to 10,000 candidates per API call with async result delivery

Improvements

Screening speed increased by 3× through updated model inference pipeline — same accuracy, less wait

Dashboard loading time reduced from 4.2s to 0.8s average on Professional plans

Research Library search now supports boolean queries and domain filtering simultaneously

Fixed

CSV export now correctly handles Unicode characters in candidate identifiers

SSO login timeout issue affecting users on Enterprise air-gapped deployments

Migration note

v2.4 is backwards compatible with all v2.x API integrations. The new batch screening endpoint is additive — existing single-candidate calls are unchanged. Benchling integration requires a one-time OAuth connection from your workspace settings.