From the source
On BigLaw Bench, GPT-5.6 Sol scored 92.7%, a step up from GPT-5.5’s 91.7%, with perfect performance across risk assessment & compliance and deal management.
There were also strong results across document review and analysis, drafting, and case management.
These are high-volume legal workflows where reliable accuracy compounds into real time savings.
On LAB, which is graded on a stricter all-pass basis, GPT-5.6 Sol scored 2.5%, showing there is room to improve on end-to-end task completion.
Additionally, early testing showed opportunities for growth in certain categories like regulatory and advising, transcript analysis, and transaction structuring — areas that tend to be challenging for frontier models.
GPT-5.6 Sol will be available in Harvey to US customers starting this week, with EU and AU availability to be announced.
Next Up
