Help
Limits and reliability
Every limit in llmwise in one place, what happens when you reach each one, and what llmwise does when a model's provider has trouble.
Limits
Message counts are on the pricing page; these are the rest.
Replies an hour
Every plan has a soft cap of 100 replies in a rolling hour, counting new messages and approved tool steps together. At the cap, wait a little: the oldest replies drop out of the hour.
Reply length
A reply can be up to 8k tokens on Free (Pro 8k, Max 16k, Ultra 16k, Studio 16k). At the limit it stops and offers Continue, which is a new message at the same price.
Long chats
Once the conversation sent to the model passes 64k tokens, each reply counts as 2; past 128k, as 4; past 200k the chat is full. The message box says so before you send: “Long chat — each reply counts as 2. Start a new chat to save messages.”
Web searches
On Free, web search works inside your 5 trial messages, up to 3 searches a day; reading a link counts as one. On a paid plan each search counts as one more message of the model in use. One message can search up to 5 times.
Files
Up to 10 files a message, each up to 20 MB; images up to 5 MB, PDFs up to 100 pages.
Code and apps
A code run stops after 60 seconds. A backend app runs 10 minutes, longer while you watch it, up to 30. Both are on paid plans.
Tool approvals
One message can go back and forth with approved tools up to 10 times; after that, send a new message.
Skills
Up to 20 skills of your own on Free, 100 on a paid plan; the built-in ones don't count.
Models on Free
Free's one-time trial of 5 messages (one of them can be on Claude Opus 5.5) works on every model but Claude Fable 5.1 and GPT-6 Astra, which need a paid plan.
Fair-use limit
On a paid plan: Pro $7.50, Max $20, Ultra $42 and Studio $85 of AI cost a billing period, at list prices; a $10 top-up adds $4 of room in the period it's bought in. At 80% we tell you; at the limit everything pauses until the period renews or you top up. Using every message on your plan at typical sizes stays under it; very large messages and heavy research use it faster.
When a count runs out
Out of everyday messages (a paid plan): the everyday models wait until the count refills at 00:00 UTC. The other models draw on the monthly allowance, so they keep working while it has messages left.
Out of monthly messages (a paid plan): the models that draw on the allowance stop until it renews or you top up. Everyday models keep working.
At the fair-use limit (a paid plan): every message pauses, the everyday models' too, until the period renews or you top up. We tell you at 80% first.
Out of Free's trial: Free is a one-time trial of 5 messages that doesn't refill, so the next message needs a paid plan.
How the counts work: plans, messages and limits.
When a provider has trouble
Claude, GPT, and Gemini go to the model's maker whenever llmwise can reach it directly. If the maker fails before the reply starts (an overload, a server error, a dropped connection), llmwise sends the same request to the same model through OpenRouter instead.
DeepSeek, Grok, Kimi, GLM, and Mistral are served through OpenRouter, and only by hosts that don't store or train on prompts; where a model has several hosts, OpenRouter moves to another when one is down.
Once a reply has started, a failure stops it: send the message again, or pick another model in the same chat.
A reply cut off because the provider failed costs nothing, however much of it arrived.
More in LLM failover, explained.
Questions
Are there API rate limits?
No. llmwise is a chat app: there's no API, so there are no request quotas or rate-limit headers to handle. What applies is your plan's message counts, on the pricing page, and the limits on this page.
Do I pay for a reply that fails?
Not when the provider failed. A reply cut off by a network error, an outage, an overload or a timeout costs nothing, however much of it arrived. A request the model rejects (a chat too long for it, a refusal) is charged like any reply once something came back.
Will llmwise answer with a different model if mine is down?
No. llmwise never switches you to a model you didn't pick. It retries the same model another way, and choosing a different one is up to you.
Claude, GPT, Gemini, DeepSeek, Grok, Kimi, GLM, and Mistral, in one chat.
See what a message costs before you send it. Free is 5 messages to try; sign in with an email link, no password or card.