Switch model tiers

Every chat message is sent to one of four model tiers, chosen from the model picker in the composer:

• fast — lowest cost, lowest latency. Available on every plan, including Free.
• standard — a stronger general-purpose model.
• premium — the strongest available model, for the hardest tasks.
• auto — picks fast/standard/premium for you, based on a simple heuristic (message length plus a few complexity markers like code blocks or "step-by-step") — not a trained classifier, so treat it as a reasonable default rather than a guaranteed-optimal choice.

What decides which tiers you can use

Your plan decides which model tiers you're allowed to call — this is checked on the server for every request, not just hidden in the UI. If a tier isn't included in your plan, the picker won't let you send with it. This applies to auto mode too: it only ever resolves to a tier your plan actually allows, falling back to a cheaper tier if your plan doesn't include the one it would otherwise pick.

Cost

Every message's exact prompt/completion token count and USD cost is recorded and debited from your organization's ledger balance in the same transaction as the reply — see How it works for the full request lifecycle, and Check your quota and add credit for how to see your remaining daily quota.