When chats run with caps
Found this helpful? Share it:
Found this helpful? Share it:
On the free tier, the cheapest models run in full and standard models run with caps - a bounded preview that keeps your daily allowance stretching across real work. Subscribing unlocks every model in full mode, with no caps. The caps are the same regardless of which model you picked 🔹
When a message runs capped, the following limits apply:
8,000 tokens of input per message (typed text + attachments + @-mentions)
32,000 tokens of total context (system prompt + conversation history + new message)
2,000 tokens of output per reply
A short agent loop: a capped run takes at most 2 tool calls and 3 back-and-forth steps, so it answers quickly rather than working a long task.
A reduced tool palette: no computer control and no browser apps. Files, web search, image generation, code, sub-agent delegation, and memory stay available within the message budget.
Capped messages still draw from the same daily usage allowance as any other action - there's no separate quota. The caps are about bounding the worst-case per-message cost so the free allowance stretches across a real preview, not about adding new charges.
It comes down to whether you have a subscription:
Not signed in: you get a two-message trial of Auto Fast (Gemma), then a prompt to sign up. This is a one-time trial, not a daily allowance, so signing up is the real unlock.
Free- the cheapest models (Gemma and the other cheap-band entries) run in full mode. Standard models run capped, so you can preview them on your daily allowance. Premium and frontier models aren't available on the free tier - upgrade to a plan to unlock them.
Subscribed- every model your plan includes runs in full mode. No caps ever apply.
A subscription lifts the caps: subscribers never see capped runs on the models their plan includes. See Plans, allowances, and credits.
Capped runs are disclosed at the moment of intent, not as a sticker on every model in the picker:
Model picker- every model wears a cost badge, from
a green $for the cheapest up through $$$$ for frontier models. The badge is the same for every viewer; it's a
fact about the model, not a judgment about you.
Composer banner- when you pick a model that will run capped on the free allowance, a small note appears above the input with an upgrade prompt. Drive, web search, and the rest of the composer stay available within the per-message budget.
In-thread wall- if a long chat fills its 32k cap, an in-thread card offers an upgrade, "Continue with Gemma" (which runs in full mode at 128k), or a fresh chat.
The context meter turns red around 85% and amber around 70%. That gives you a couple of turns to wrap up or switch to Gemma before hitting the cap.
Two walls can fire:
Context wall- the chat has filled its 32k context cap. You get three options: upgrade to a plan, continue with Gemma (keeps the full conversation in a 128k window in full mode), or start a new chat.
Allowance wall- your daily usage allowance is spent. Options are upgrade to a plan or wait for the allowance to refill.
Visitors who aren't signed in don't see "Continue with Gemma": they only have the trial model anyway, and the sign-up path is the real unlock.
Subscribers who run out of included usage hit a different wall: a plan wall with a pay-as-you-go option instead of caps. See Credits and auto top-up.
Related articles
Plans, allowances, and credits
How the free daily allowance, subscription plans, and Usage Credits fit together, and what happens when you reach a limit.
Credits and auto top-up
Add a prepaid Credit balance to your subscription, keep it funded automatically, and decide whether extra usage may overflow into it.
Was this helpful?