Hi @Ninh_Nguyen Thank you so much for the post! There’s no way to manually move usage between pools, but the effect you’re after already happens automatically, so there’s nothing you need to do.
Included usage is split into two pools that are tracked separately:
Cursor Models: Cursor Grok 4.5 and Composer 2.5
Other Models: third-party models, including gpt-5.6-luna-xhigh
Once your Cursor Models pool is fully consumed, Cursor Models usage starts drawing from whatever is left in your Other Models pool. So that allocation isn’t wasted, your Grok usage will pick it up after the Cursor Models pool runs out. If both pools are used up, you can keep going with on-demand usage if it’s enabled for your team.
It doesn’t work in the other direction. Third-party models can’t draw from the Cursor Models pool, so if you use up Other Models first, those models stop at the limit or move to on-demand.