How to Use AIHubMix Free Models in JanitorAI: Proxy Setup and Common Errors

JanitorAI can talk to any OpenAI-compatible API through its Proxy option, and AIHubMix is one. The whole setup is four values typed into one form:

Name:       anything you like, e.g. AIHubMix MiMo
Proxy URL:  https://aihubmix.com/v1/chat/completions
API key:    your AIHubMix key
Model:      xiaomi-mimo-v2.6-pro-free

Leave Custom prompt empty, press Test, then Add, and refresh the page. That's it.

Most failures reported in the JanitorAI community come from one of those four fields being slightly off, or from the free tier running out sooner than expected. This guide walks through the setup once, then covers each error you're likely to see and what to do about it.

Before you start

You need three things:

1. An AIHubMix account and an API key. Sign up, then create a key on the AIHubMix API key page. Copy it somewhere safe right away.
2. A bot that allows proxies. A proxy only works with bots that show "proxy allowed" above the persona selector or the start-chat button, as this community proxy guide explains. Bots with visible character definitions usually allow it.
3. A model ID. Pick one from the AIHubMix free models list and copy its ID exactly. The community's current favourite for roleplay is xiaomi-mimo-v2.6-pro-free; the model guide in this series compares the options.

One thing to know before your first message: a new account gets 10 trial calls in total, shared across all free models. After a one-time top-up of $1 or more, the account switches to daily quotas (100 requests a day, 10 a minute, 1M tokens a day). The limits guide in this series explains how far that goes in a long chat.

The four values, and where people slip

The model ID goes in the Model field. It is lowercase and keeps its -free suffix. "Xiaomi Mimo V2.6 Pro (free)" is the display name and won't work; xiaomi-mimo-v2.6-pro-free will.

\1

Next to the Proxy URL label there's an Add /chat/completions button. It only appends /chat/completions to whatever is already in the field. If you pasted the full URL above, don't click it, or the path ends up twice. If you pasted https://aihubmix.com/v1, one click completes it.

The API key field says "(optional)". Fill it in anyway: without a key, every request is rejected.

Custom prompt overrides the system prompt for this proxy and is sent at the top of every request while the proxy is active. AIHubMix doesn't need one, so leave it empty.

Set it up in JanitorAI

JanitorAI redesigned its chat settings on 19 July 2026, so older videos show different screens. The steps below were checked against the current web version in October 2026:

1. Open a chat with a proxy-allowed bot.
2. Open the menu at the top right and choose Settings.
3. Under Provider, choose Proxy.
4. Under Proxy Configurations, click + New.
5. In the New Configuration form, fill in Name, Proxy URL, API key and Model with the values above. Leave Custom prompt empty.
6. Click Test. If the setup works, nothing pops up; a message appears only when something is wrong.
7. Click Add, then refresh the page or restart the chat.

The refresh in step 7 is the step people skip most. Until the page reloads, JanitorAI can keep sending messages through the old setting.

You can save several configurations, for example one per model, which makes switching quick when one model is busy. To add a backup, open a saved configuration, click Duplicate, and change only the Model field.

Settings for your first chat

Open the menu → Settings again and look at the generation controls. Two of them matter for a first run, according to Arcanum RPGs' walkthrough of the current panel:

• Context size. With an outside model the slider goes up to 128,000 tokens. Start between 16,000 and 32,000. JanitorAI's own tooltip says to lower it if replies fail to generate.
• Max tokens. 0 lets the model choose the reply length. Any other number is a hard stop, so a low value can cut replies off mid-sentence.

Press Save; nothing applies until you do.

Context size matters for more than quality. Every turn sends the whole context again, and all free models share one pool of 1M tokens per day. Once a chat fills a 32,000-token context, that pool covers roughly 30 replies, well under the 100-request cap. The limits guide in this series works through the numbers.

Optional: test your key outside JanitorAI

If you're comfortable with a terminal, this request separates "my key is wrong" from "my JanitorAI setup is wrong". It uses the same endpoint the AIHubMix quick start documents, and it uses up one request (one of your 10 trial calls if you haven't topped up).

export AIHUBMIX_API_KEY="paste-your-key-here"

curl https://aihubmix.com/v1/chat/completions \
  -H "Authorization: Bearer $AIHUBMIX_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "xiaomi-mimo-v2.6-pro-free", "messages": [{"role": "user", "content": "Say hello in five words."}]}'

A reply with text in it means the key and model ID are fine, and the problem is in JanitorAI's fields or the missing refresh.

Common errors and how to fix them

"Network error" or "Failed to fetch"

Test shows "Network error. Try again later!", or the chat shows "A network error occurred, you may be rate limited or having connection issues: Failed to fetch".

• Why: The request never reached AIHubMix, or the reply never made it back to your browser. Despite the wording, "Failed to fetch" points to a connection problem, not a quota. The usual cause is a VPN or proxy that blocks or misroutes aihubmix.com; a browser extension that filters requests can do the same. In AIHubMix's own test run of this guide, a VPN was the cause.
• Fix: Turn the VPN off, switch to another server, or make sure it lets aihubmix.com through. Pause ad-blocking or privacy extensions on janitorai.com. Then press Test again.

Test shows an error, or the bot never replies

• Why: The Proxy URL is incomplete (base URL or homepage only), has a typo, or ends in /chat/completions twice because the Add /chat/completions button was clicked after pasting the full URL. AIHubMix answers a wrong path with "The requested API endpoint does not exist."
• Fix: Paste https://aihubmix.com/v1/chat/completions exactly, with no trailing space.

"Model not found" or a model error

• Why: The Model field holds a display name, has a capital letter, or lost its -free suffix.
• Fix: Copy the ID from the free models list again and paste it as-is.

"Invalid key" or an authentication error

• Why: The key belongs to another service, was deleted, has a stray space, or the field is empty. AIHubMix returns "no key provided" for an empty key and "invalid key" for a wrong one.
• Fix: Create a fresh key on the AIHubMix key page and paste it again.

You changed the settings but nothing changed

• Why: JanitorAI is still using the configuration it loaded before.
• Fix: Refresh the page or restart the chat, and check that the right configuration is selected.

The bot's "reply" is a notice that your trial is used up

• Why: The 10 trial calls are gone. AIHubMix keeps answering with a note instead of an error, so in JanitorAI it most likely shows up as a message in the chat.
• Fix: Top up $1 or more once; the account switches to daily quotas for free models. The top-up guide in this series covers the steps.

Error 429

A 429 has two common causes:

• You hit a cap. Past your per-minute or per-day limit, the API returns 429 until the next minute or day starts. Some models have lower caps than the standard 100 a day and 10 a minute (for example, qwen3.6-plus-preview-free allows 50 a day and 5 a minute), and the shared 1M-token pool can run out first in long chats.
• The model is busy. Free capacity on some models is limited, and the AIHubMix changelog has warned that 429 errors "may occur" on a free model for that reason. When many people pile onto one popular model at once, this is the usual cause.

Fix: Wait a minute and retry, send messages less often, or switch to a second saved configuration with a different model. If the daily cap is the problem, the quota resets once a day.

A 404 that mentions model_retired

• Why: The model has been removed. Since 1 September 2026, retired models return an explicit model_retired 404.
• Fix: Pick another model from the free list and update the Model field.

Free models stopped working after you used paid ones

• Why: When an account balance falls below -$1, free models can't be called until the account is topped up again.
• Fix: Top up to bring the balance back above that line.

Still stuck?

The AIHubMix contact page lists the official Discord server, the support email (feedback@aihubmix.com), and the status page. The status page tracks services such as Chat Completions, not individual models, so a busy model won't show up there.

To get a fast answer, include:

• the request time and your time zone
• the model ID and the endpoint (the Proxy URL)
• the HTTP status or error text, and a trace ID if one was shown
• what you expected and what happened instead

Leave out your API key. Support doesn't need it, and anyone who has it can spend your quota.

FAQ

Is AIHubMix free to use in JanitorAI?
Every new account gets 10 trial calls across all free models. After a one-time top-up of $1 or more, free models switch to daily quotas: 100 requests a day, 10 a minute, and 1M tokens a day shared across them.

What exactly goes in the Proxy URL field?
https://aihubmix.com/v1/chat/completions, the full address. If you paste only https://aihubmix.com/v1, click the Add /chat/completions button once to complete it.

Why don't I see the proxy option for some bots?
Only bots marked "proxy allowed" work with a proxy. Other bots stay on JanitorAI's own model.

Which model ID should I start with?
Any ID from the free models list works if copied exactly. Many JanitorAI users start with xiaomi-mimo-v2.6-pro-free; the model guide in this series compares the alternatives.

Do I need to refresh after changing the model?
Yes. Save the configuration, then refresh the page or restart the chat so JanitorAI picks up the change.

Why do I keep getting 429 errors?
Either you passed a per-minute or per-day cap, or the model is busy. Wait a minute, slow down, or switch to another saved configuration with a different model.

Can I share my API key with friends?
No. The Terms of Service bar sharing an AIHubMix account, and anyone using your key spends your daily quota.

Keep reading: the AIHubMix for JanitorAI series

• Your setup works, but how many messages will it actually give you per day? The token math is here: AIHubMix Free Model Limits for JanitorAI: 10 Trial Calls, the $1 Unlock, and the Token Cap
• If the trial notice already showed up in your chat, this walks through unlocking the daily quotas: How to Pay on AIHubMix with Alipay
• Not sure the model ID you pasted is the right one for roleplay? Compare the free options: Which AIHubMix Free Model to Use for JanitorAI Roleplay
• Before your first long chat, see how zero data retention keeps your messages from being stored on AIHubMix: AIHubMix Supports JanitorAI Roleplay Chat, with Zero Data Retention on by Default

Sources

• Free AI Models (AIHubMix)
• Quick Start (AIHubMix Docs)
• Changelog (AIHubMix Docs)
• custom-proxy-for-janitorai (GitHub)
• Janitor AI Generation Settings (2026) (Arcanum RPGs)