Skip to main content

Model control

Model control is where organization admins choose which models and providers your org can use, and set defaults for each surface — desktop, CLI, cloud agents, and mobile.

Members only see models you allow. If a model is disabled, it won't appear in pickers — and manual picks in the CLI are blocked too. This is not where you change your commercial plan; your contract defines the ceiling of what can be enabled. Model control is how you curate that catalog for day-to-day work.

What this page is for

You want to…Model control helps you…
Keep spend predictableEnable a focused catalog instead of every model
Onboard new teams quicklySet sensible defaults on desktop, CLI, cloud, and mobile
Separate fast vs deep workAllow a standard set for daily coding and reasoning for reviews
Lock down productionNarrow models on production projects; looser on sandbox
Fix "model not allowed"Trace org, team, or project limits — see When a model is blocked
Undo a bad rolloutUse Change history to restore a prior catalog version
Model control vs Console vs Usage

Model control curates which models members can use inside your plan. Included providers and model groups come from your agreement — set in the Console, not here. Allowed models still consume allowance — pair catalog choices with Usage & billing budgets and weekly review during rollout.

Who can see and change what

AreaTypical access
Org-wide catalog (providers, models, defaults)Admin or super admin
Team or project model limitsAdmin; delegated managers where your plan allows
See which models they can useAny member (pickers show allowed models only)
Override a blocked model one-offNot supported — fix catalog or project scope instead
Change history / restore prior versionAdmin

If you cannot open Model control, your role may be member only — ask an admin in Members. Members still pick from allowed models when they run agents; they just cannot change org policy.

Why model control matters

Without curationWith model control
Everyone picks any modelAdmins enable a focused catalog
Spend varies wildlyFewer, well-chosen models → predictable usage
Onboarding is confusingSensible defaults on every surface
Experimental models in prodStricter limits on production projects

Model control works best together with Usage & billing — allowed models still consume allowance when people run agents. Tightening the catalog is often the fastest cost lever before raising budget caps.

How model control fits together

Think of Model control as the model catalog boundary; other modules handle spend and people:

ModuleWhat it affectsTied to Model control
UsageSpend and denialsPolicy blocks show on Usage — no spend, but explains failures
SettingsOrg structureTeams for squad-level narrowing
MembersRolesWho can change org-wide catalog
ProjectsWork groupingProject-level model limits
Console / ContractsPlan ceilingWhich providers and groups you can enable at all

When someone reports a blocked run, check model policy here and budget policy on Usage — both can stop runs, but only model usage drives spend.

In practice

You want to…Do this
Roll out X4RGE to a new teamEnable one standard set of models + defaults
Cut costDisable expensive models; keep a fast default for daily work
Support deep reviewsAllow a reasoning tier for leads; standard for everyone else
Different defaults per toolSet desktop vs CLI vs cloud defaults separately
Lock down productionNarrow models on production projects; looser on sandbox
Fix "model not allowed"See When a model is blocked
Watch impact after a changeReview Usage trends and denials for a week

What admins control

AreaWhat it means
ProvidersWhich model providers are on for your org (within your plan)
ModelsWhich specific models or model groups members can use
Per-surface defaultsWhat model loads when someone doesn't pick one
Team / project limitsOptional narrower lists for a squad or repo (cannot exceed org catalog)

Only admins and super admins change org-wide catalog settings. Delegated managers may narrow policy for their team or project where your plan supports it — never widen beyond the org catalog or plan.

Set org-wide catalog

  1. See what your plan includes

    Your contract defines the ceiling — you can only enable providers and models your plan includes. Unavailable entries appear disabled in the dashboard. If something is missing, see Contracts & capacity or Help.

  2. Enable what you use in production

    Turn on the providers your team actually uses. Prefer model groups (for example "standard" or "reasoning") over long manual lists so new models in a group roll out smoothly.

    ApproachGood for
    One standard groupMost engineering teams — predictable cost
    Fast + reasoning splitCheap triage vs deep design work
    Block one costly modelKeep a group enabled but deny a single outlier
  3. Set defaults and save

    Pick defaults for each surface (below), save, and tell the team which models to use for which jobs. Changes are recorded in the audit log.

Start small

Enable a small catalog first, watch Usage for a week, then expand if you need more capability. Most teams need fewer models than they expect at rollout.

Defaults per surface

Defaults improve onboarding — members get a good experience without configuring anything.

SurfaceTypical default
DesktopBalanced coding model your developers prefer
CLISame as desktop, or a faster model for scripts and CI
CloudModel suited to longer autonomous runs
MobileSmaller, faster model for quick questions

You can use the same default everywhere or tune per surface for cost and quality. The CLI default applies to headless and automation when no model is specified.

Team and project limits

Delegated managers may narrow org policy for their team or project — for example, production projects allow only standard models while a sandbox allows experimental ones.

They cannot enable models the org catalog blocked or your plan excludes. Org policy is the outer boundary; team and project settings refine it inward.

Cost lever

Restricting the catalog is one of the fastest ways to keep spend predictable. Pair model control with budget policies and review usage weekly during rollout.

When a model is blocked

If someone reports "model not allowed" in the desktop app or CLI:

  1. Confirm which model and surface

    Note the model name and where they ran — desktop, CLI, or cloud.

  2. Check the org catalog

    In Model control, see whether that model (or its group) is enabled for the org and surface.

  3. Check team or project scope

    If they're on a restricted project, they may need an allowed default or a catalog update at project level — not a one-off workaround.

  4. Fix or redirect

    Either add the model to the allowed set, adjust the project limit, or ask them to use the org default for that surface.

Policy denials may also appear on the Usage page when someone tries a blocked model — useful when the member only saw a generic error in the app.

Change history

If a recent catalog change caused confusion, use Change history in Model control to restore a prior version instead of re-clicking every model manually.

Questions about model control?

QuestionAnswer
Why can't I open Model control?You may need admin — ask in Members
A model is grayed out in the dashboardYour plan may not include it — Contracts
Member can't pick a model they used yesterdayCheck org catalog, then team/project limits; see Change history
Run blocked but Usage shows no spendLikely model policy, not budget — check denials on Usage
CLI --model failsCLI validates against the CLI surface catalog — may differ from desktop
Who changed the catalog?Org audit log — see Governance for audit context
Should we allow every model in the plan?Usually no — start with one standard group and expand from Usage data