SayTwo teammates both say they are using Grok. One is directing apps on a cloud computer; the other is editing code in Agent. Why do their controls and bills differ?
DoName the product before the model. The first teammate is using Grok BotAn early AI teammate product from SpaceXAI and Cursor that works across apps on a persistent cloud computer and asks for approval on selected actions; launch materials describe it as beta, but Cursor says Beta Services terms do not apply. Press Enter for the full definition.; the second selected the Grok 4.6A Cursor and SpaceXAI frontier model for complex coding and knowledge work; it is a selectable model and is separate from the Grok Bot product. Press Enter for the full definition. model in a Cursor coding surface.
SeeThe largest difference is now explained without blaming model quality: separate products, separate controls.
DoFor the coding user, record both monthly pool balances and the plan. Grok 4.6A Cursor and SpaceXAI frontier model for complex coding and knowledge work; it is a selectable model and is separate from the Grok Bot product. Press Enter for the full definition. sits with Grok 4.5 and Composer 2.5The current Composer release, better at long-running tasks and at judging when a job needs a light touch versus deep work. Press Enter for the full definition. in Cursor ModelsCursor's plan and usage pool for models it labels as Cursor Models, kept separate from the third-party Other Models pool. Press Enter for the full definition.; a Start user cannot switch into Other ModelsCursor's pool of supported third-party models from providers such as Anthropic, Google and OpenAI, with separate plan access and usage from Cursor Models. Press Enter for the full definition., Auto or paid overflow.
DoIf the coding user is on Teams or Enterprise Auto, capture the mode. Cost keeps the prior Auto logic and uses its fixed rates. Balance and Intelligence use Cursor Router's classifier, so read the serving model from usage instead of treating Router as one named model.
SeeThe route is now reproducible: plan, pool and optimization mode are visible, with the serving model recorded for Balance or Intelligence.
DoIf someone proposes Max, label it correctly: it belongs to legacy request-based plans and costs the model API rate plus 20%.
SeeMax is now recorded as a legacy billing path, not a universal quality switch.
SayI can now compare the two sessions without calling either one broken.
SeeThe report names the product, model, pool, plan and route. Nobody treats a plan boundary or a different Grok product as a regression.