Thank you for the update!
Additional observation: it post thougths into the model transcript even if you set ThinkingConfig(thinking_budget=1024, include_thoughts=False)
Please consider keeping half cascade alive for a few more months.
1 Like