A new AIHubMix account gets 10 trial calls in total, shared across every free model. They are not 10 per day or one every few minutes. A one-time top-up of $1 or more switches the account to daily quotas: 100 requests a day, 10 a minute, and 1M tokens a day, shared across the whole free catalog. For long JanitorAI chats, the 1M-token pool is usually what runs out first, well before you reach 100 replies.
The rest of this guide explains each limit, shows how fast a JanitorAI chat uses the token pool, and explains what a 429 error is telling you. All figures come from the AIHubMix free models page as of October 12, 2026.
| Account stage | What you get | Expiry |
|---|---|---|
| Before any top-up | 10 calls in total, all free models | Never expires |
| After a top-up of $1 or more | 100 requests/day, 10/min, 1M tokens/day | Resets every day |
Stage one: 10 trial calls, then a note instead of an error
The trial is a single bucket of 10 calls. It doesn't refill, and calls to different free models all come out of the same 10. In JanitorAI, each reply you ask for uses one call. Pressing regenerate asks for another reply, so it uses another call too.
When the 10 calls are gone, the API doesn't fail. In the free models page's words, it keeps answering "with a note that the trial is used up — your integration never sees a hard error." JanitorAI will probably show that note as if the character had said it. If your character suddenly starts talking about a trial or a quota, the trial has run out.
The trial is kept small to prevent abuse: enough calls to try the service, then a small one-time top-up before regular daily use.
Stage two: daily quotas after a one-time $1 top-up
According to the page, "any top-up of $1 or more permanently switches your account to daily quotas." It is a one-time step, not a subscription, and the quotas have no expiry. How to pay is covered in the top-up guide in this series (linked at the end).
After that switch, you get:
- 10 requests per minute. This cap stops rapid back-and-forth and quick rerolls.
- 100 requests per day. This is the ceiling on replies per day.
- 1M tokens per day. This is one pool for every free model. Changing models does not give you a fresh million.
The page's FAQ describes all three limits as "shared across the free catalog." It says they reset every day, but gives no reset hour or time zone. If the exact time matters to you, ask support (details at the end).
Some models count as several requests
A few free models have lower caps because, as the page puts it, they "carry a higher request weight — one call counts as several requests."
| Model | Per day | Per minute |
|---|---|---|
| qwen3.6-plus-preview-free | 50 | 5 |
| coding-glm-5-turbo-free | 50 | 5 |
| k2.6-code-preview-free | 50 | 5 |
Caps apply after the top-up. Every other free chat model in the table is listed at 100 per day and 10 per minute. The table also lists lower caps for several GPT and Gemini free models, but their model pages now say they have been retired, so they are left out here.
Why tokens run out before requests in JanitorAI
On every turn, JanitorAI sends the model the character card, your persona, and as much of the chat history as fits in the context size you set. As the Arcanum RPGs guide to Janitor's generation settings puts it: "Every turn sends the whole context again." With an outside model, Janitor's context slider goes up to 128,000 tokens. Once a chat is long enough to fill that window, every reply costs roughly the full context size in tokens.
The table below estimates how many replies the 1M pool lasts at common context sizes. Assumptions:
- the chat is long enough to fill the context on every turn;
- each reply is about 500 tokens;
- both the tokens sent and the tokens received count toward the daily pool;
- you don't use free models anywhere else that day.
| Context size | Tokens per reply | Replies per 1M | Limit you hit first |
|---|---|---|---|
| 8,000 | ~8,500 | ~117 | 100 requests |
| 16,000 | ~16,500 | ~60 | 1M tokens |
| 32,000 | ~32,500 | ~30 | 1M tokens |
| 64,000 | ~64,500 | ~15 | 1M tokens |
| 128,000 | ~128,500 | ~7 | 1M tokens |
Each row is 1,000,000 divided by (context size + 500).
At Janitor's maximum context, the free tier lasts about seven turns of a long chat. At 16,000 it lasts about 60. The settings that move this number:
- Context size. Arcanum's guide suggests starting between 16,000 and 32,000, and Janitor's own tooltip says to "lower it if replies fail to generate." Halving the context roughly doubles your replies per day.
- Max tokens. In Janitor, 0 "lets the model choose the length." A non-zero value is a hard stop, which keeps replies shorter but can cut them off mid-sentence.
- Reasoning models. MiMo V2.6 Pro lists "Thinking" among its capabilities. Thinking output is still output, so a reasoning model may use more of the pool per reply than a model that answers directly. This is an estimate, not a measured figure. If the pool drains faster than the table suggests, a non-reasoning model is worth trying.
Switching models doesn't refill the pool, because every free model draws from the same 1M tokens.
What a 429 error means
The free models page says: "Past a cap the API returns 429 until the next minute or day starts." A 429 has one of two causes.
You hit your own cap. You sent more than 10 requests in a minute, or you've used up the day's requests or tokens. Waiting is the only fix: a minute for the per-minute cap, the next daily reset for the others. Sending messages more slowly, and not mashing regenerate, keeps you under the per-minute cap.
The model itself is busy or capped. Some free models run on limited capacity. The AIHubMix changelog has said of a free model: "Resources are limited, stability is not guaranteed, and 429 errors may occur." In that case, waiting a bit or switching to another model helps. JanitorAI lets you save several proxy configurations, so keeping a second one with a different model makes switching quick.
Note that the AIHubMix status page tracks services such as Chat Completions, not individual free models. A green status page doesn't rule out a busy model.
One more rule: a balance below -$1
From the changelog entry of June 23, 2026: "When an account balance is below -$1, free models can no longer be called until the account is topped up." This only matters if you also use paid models on the same account and let the balance go negative. Free use alone doesn't move the balance.
When replies stop: a quick check
- The character's reply is a notice about a trial being used up: your 10 trial calls are gone, so top up $1 or more.
- 429 after several quick messages or rerolls: you hit the per-minute cap, so wait a minute.
- 429 after a long session: you used up the day's requests or tokens, so wait for the daily reset or lower the context size.
- 429 that persists after waiting a minute, with quota left: the model may be busy, so switch to a saved configuration with another model.
- Free models stopped right after using paid models: check whether your balance is below -$1.
If a limit looks wrong, contact AIHubMix support with the request time and time zone, the model ID, and the HTTP status. Leave out your API key; support doesn't need it.
FAQ
Are the 10 trial calls per day?
No. They are 10 calls in total, shared across all free models, and they never expire. Daily quotas only start after a top-up of $1 or more.
Do I have to top up every month to keep the daily quota?
No. The free models page says a top-up of $1 or more permanently switches the account to daily quotas, which have no expiry.
Does each free model get its own 100 requests and 1M tokens?
No. The page describes the limits as shared across the free catalog, and all free models draw from one 1M-token pool. Some models also count one call as several requests.
Why did I run out after about 30 messages?
With a long chat and a context size around 32,000 tokens, each reply sends roughly 32,000 tokens. That uses up the 1M daily pool after about 30 replies. Lowering the context size gives you more replies.
When does the daily quota reset?
The free models page says quotas reset every day but doesn't publish the hour or time zone. Contact support if you need the exact time.
What does a 429 error mean?
Either you passed your per-minute or daily cap, or the model you picked is busy. Wait a minute or until the reset for the first; switch to another free model for the second.
Can I see whether a free model is overloaded?
Not on the status page, which tracks services rather than individual models.
Keep reading: the AIHubMix for JanitorAI series
- Not connected yet? The limits only matter once JanitorAI is talking to AIHubMix: How to Use AIHubMix Free Models in JanitorAI: Proxy Setup and Common Errors
- Ready to move past the 10 trial calls? The step-by-step payment guide: How to Pay on AIHubMix with Alipay
- Choosing a model with the full 100-a-day cap, and keeping a backup for 429s: Which AIHubMix Free Model to Use for JanitorAI Roleplay
- What AIHubMix records about those requests, and what it doesn't: AIHubMix Supports JanitorAI Roleplay Chat, with Zero Data Retention on by Default
Sources
- Free AI Models (AIHubMix)
- Xiaomi MiMo V2.6 Pro (free) (AIHubMix)
- Changelog (AIHubMix Docs)
- Janitor AI Generation Settings (2026) (Arcanum RPGs)
- Contact AIHubMix (AIHubMix)
