AI that answers faster than you read
Paid VoidBox servers run on GLM 5.3 Flash: it reads text, images and video and holds a 1M-token context. Ultra fast mode sends every reply to whichever provider is answering quickest right now — reaching up to 100 tokens a second.
An AI Discord bot with a personality and a memory
- No API key, no setup (free): add VoidBox and mention it. Free servers run on free models, so it costs you nothing.
- It remembers you (free): your last few exchanges, for 7 days, in any server. See it with
/ai memory; wipe it any time. - Tones: normal, Cry Baby Mode (free), Savage (Premium) and Unhinged (Premium+, age-restricted channels only). Every tone stays away from sensitive topics.
- Your own AI personality (Premium+): give the bot a custom personality and backstory for your server.
- It can see images and GIFs (Premium): post a picture or a meme and it answers about it.
Games it built from one prompt
Ask VoidBox to build something on Premium and it writes the whole thing, then test-runs it and fixes it before you get the file. Both games below were written by GLM 5.3 Flash from a single message.
Which AI models VoidBox uses
| Used for | Model | Plan | Speed |
|---|---|---|---|
| Chat on free servers | dots 3 Note (free community model) | Free | … |
| Everyday chat | DeepSeek V4 Flash | Premium | … |
| Images, video and building things | GLM 5.3 Flash (1M-token context) | Premium | … |
| Hard questions (reasoning) | gpt-oss-120b | Premium | … |
| Roleplay | Dolphin Mistral 24B Venice | Premium | … |
| AI moderation | gpt-oss-20b | Premium | … |
| Image generation | Meta Muse | Premium | Pictures, not text |
| Boombox Evil persona | L3.3 Euryale 70B | Premium+ | … |
Speed is in tokens a second: first the average of VoidBox's real replies over the last 7 days (counted from sending the question, so it includes the wait for the first word), then the fastest provider right now, which Ultra fast mode aims for.
Routing that measures, not guesses
Provider speed changes hour to hour — the same model can stream at 120 tokens a second on one host and 40 on another. Ultra fast mode keeps its own speed measurements from real replies, ranks the providers by them, prefers the fastest, and drops a provider that keeps failing. When one is slow or busy, the next reply goes elsewhere automatically. It costs nothing extra.
“Up to” is deliberate: 100 tokens a second is the measured ceiling, not a promise for every reply.
Private by default
AI requests go to providers with zero data retention first; only if none is available may a provider keep a prompt for a while. Your members' messages are never used for training.
Free servers get AI too
AI chat is not paid-only. A free server gets 50 replies a day, counted per server, on a free community model — best effort, with no uptime guarantee. Paid plans replace it with GLM 5.3 Flash and remove the daily cap.
What paid plans add
- Premium ($4.99) — GLM 5.3 Flash from a 4M-token monthly pool, vision, image generation, AI moderation, a reasoning model, live web search and roleplay.
- Premium+ ($9.99) — a 10M-token pool, plus DM chat and the Boombox Evil persona.
Questions
Is there a free AI Discord bot with no API key?
Yes. VoidBox works as soon as you add it: mention it and it answers, using free models, with nothing to set up.
Can a Discord AI bot have a custom personality?
Yes. VoidBox has tones like Savage and Cry Baby Mode, and Premium+ lets you write your own personality and backstory for your server.
Does the VoidBox AI remember conversations?
Yes, your last few exchanges for 7 days, across servers. You can see or wipe it with /ai memory.
Can a Discord AI bot see images?
On Premium, VoidBox reads images and GIFs you post and answers about them.
What is the fastest AI Discord bot?
VoidBox's Ultra fast mode streams GLM 5.3 Flash replies at up to 100 tokens a second by routing every reply to whichever provider is answering quickest, based on its own live measurements.
Which AI model does VoidBox use?
Paid servers use GLM 5.3 Flash, which reads text, images and video and has a 1M-token context. Free servers use a free community model for up to 50 replies a day.
Does the VoidBox AI store my messages?
VoidBox sends AI requests to zero-data-retention providers first, which keep nothing. Only if none is available may a provider keep the prompt for a while. Messages are never used for training.
Why does free AI sometimes say it is busy?
Free servers run on free community AI models that are shared by many apps at once. They cost nothing, which is why they can be offered for free, but at peak times they can be full and a reply fails. Trying again a minute later usually works. There is also one daily allowance of free AI requests shared by every free server on VoidBox, reset at midnight UTC; on a very busy day it can run out before a server reaches its own 50, and the bot says so. Premium and Premium+ are never affected by any of this: they run on paid models with their own capacity, so paid AI keeps working normally even when free AI is busy or out for the day. Each paid model is served by 20 to 30+ different providers, and a reply automatically moves to another one if a provider is slow or down, so a paid reply failing is very unlikely.
Is AI free on VoidBox?
Yes, in part: every free server gets 50 AI replies a day on a free model. Premium and Premium+ add GLM 5.3 Flash with no daily cap.