vust

AI CHAT

Start the chat here. Keep one thread.

Two real answers in the browser before sign-in. Connect Telegram to save the same thread; open the current model catalog in @vustbot when you want to switch.

Free · 2 replies before connectionPro/Max · full catalog in @vustbot
  • 2 replies before sign-in
  • Context for one follow-up
  • No personal API key
Ask the live model
Real answers and one contextual follow-up — right here
Web preview · vust Fast
Ask your first question
The selected model stays active. Your follow-up keeps compact context.
vust Fast · GPT-6 Luna
Enter sends · Shift+Enter adds a lineNo replies used yet
Guest chat only · kept 24 h · not saved to an account

THE LINEUP

Current models, one familiar chat

Start with vust Fast here. In @vustbot, the current catalog adds everyday and flagship models with explicit access before anything is sent or charged.

Every delivered answer names the model that actually responded.

22 current models
A few everyday and flagship choices
22 models

Everyday · pre-flagship

  • vust Fast

    GPT-6 Luna

  • vust Smart

    Gemini 3 Flash

  • vust Deep

    GPT-6.1 Terra

  • vust Claude

    Claude Sonnet 5.5

Flagships

  • vust Max

    Claude Opus 5.5

  • vust Long

    Gemini 3.1 Pro

  • vust GLM

    GLM-5.3

  • vust Grok

    Grok 4.7

Specimens

See the difference

What one conversation looks like in @vustbot, where the models are switchable.

Compare two models, one question

You ask once

Which database fits a 50M-row analytics workload — Postgres or ClickHouse?

Two answers, same thread

GPT-6.1 Terra weighs operational simplicity; Claude Sonnet 5.5 stresses columnar scan speed. In @vustbot you re-ask with another model in the same thread and keep the answer that fits.

Free everyday asks

Simple request

Rewrite this email so it sounds friendlier but still professional.

Fast model

Handled in seconds by a fast model — no card needed to start. Strong models stay reserved for the questions that actually need them.

frontier access

The blocker

ChatGPT Plus won't accept my card, and I'd rather not install anything extra just to pay.

plans paid by card inside Telegram

Start with a free first answer, pay by card inside Telegram, and reach GPT, Claude, Gemini and Grok from the same Telegram you already have open — no Western card subscription, no separate website sign-up.

Rehearse the conversation you keep putting off

You have rehearsed it in the shower twenty times, and every version ends with you saying yes to something you did not want. A real rehearsal is different: you say the words out loud to a counterpart who pushes back, and you hear which sentence falls apart before it matters.

Describe the situation to @vustbot, say what outcome you want and what you will not agree to, and ask it to play the other person. Practice one reply at a time. The other side's lines are a realistic guess, not a prediction, so practice more than one way they might react.

Practice a conversation you are afraid of

This is an illustrative rehearsal.

The brief: "I need to tell my manager I am overloaded. Outcome: one of my three projects moves to someone else or gets a later date. Boundary: I will not keep working weekends."

Manager (played): "Everyone is stretched right now. Can you hold on until the end of the quarter?"

Your first attempt: "I guess I can try."

Your revised reply: "I can hold the two main projects to their dates. The third one will slip by a month unless it moves to someone else. Which would you prefer?"

Check: the revised reply offers a choice between two real options and never says "I'll try". The weekend boundary is never traded away.

Handle a difficult client without giving away the store

This is an illustrative rehearsal.

The facts: the contract includes two rounds of revisions. The client wants a fourth round, free, and says the work "missed the brief".

Client (played): "This should have been right the first time. I expect the changes at no cost."

Your reply: "I want this to work for you. The two included rounds are done, and I can do a third round at the revision rate in the contract. If something specific missed the brief, send me the line from the brief and I will fix that part at no charge."

Check: facts are separated from interpretation. "Missed the brief" becomes a question about a specific line, not an argument about quality, and the reply makes no new free promise.

Write the message you should not send

Sometimes the useful thing is to write the furious message in full, and then not send it. This is an illustrative exercise.

The unsent draft: "You have ignored three emails about the broken heating. This is the last time I ask nicely."

What it is for: say everything here. Then ask @vustbot to split it into what is release and what is communication.

What can be sent: "The heating has been broken since 2 November. I reported it on the 2nd, 6th and 10th. Please confirm a repair date by Friday."

Decision checklist: hold the draft overnight, send the factual version, or send nothing. Nothing leaves the chat unless you copy it out yourself. If a draft contains a threat, it stays unsent.

Practice holding a personal boundary

This is an illustrative rehearsal.

The situation: a relative comments on how you raise your children at every family dinner.

Your boundary sentence: "I know you care about the kids. I am not going to discuss how we raise them at dinner."

Relative (played): "I'm only trying to help. You're so sensitive."

Your second reply: "I hear that you mean well. My answer is still the same. Let's talk about something else."

Your exit: "I'm going to step out for a few minutes," said once, calmly, if the topic comes back a third time.

Check: every line describes what you will do, not what is wrong with the other person. If a conversation involves threats or your safety, contact local support services instead of rehearsing it.

See the gap between what you meant and what they heard

This is an illustrative example.

What you wrote: "Fine, do it your way."

What you meant: "I agree with your plan."

Plausible readings: agreement, reluctant agreement, or giving up in annoyance. A short sentence starting with "Fine" carries all three, and the reader picks one based on their mood, not yours.

A clearer version: "Your plan makes sense. Go ahead, and tell me if you want a second pair of eyes on the timeline."

A clarifying question when a message already landed badly: "I think my 'fine' came across wrong. I meant I agree. Did it read differently to you?"

Check: the readings are labeled as possibilities. Nobody can know what the other person thought, and the clarifying question is how you find out.

Open @vustbot, describe the conversation you are dreading, and ask it to play the other side.

Find business ideas in years of your own notes

Your notes are full of complaints you wrote down and forgot: a tool that annoyed you, a task you did by hand every month, a question three friends asked you. Each one is small. Together they are a map of problems people actually have, which is where good business ideas come from.

Paste a handful of note excerpts into @vustbot and ask it to pull out problems, not product names. For each problem you get who has it, when it happens, and how often it appears in your notes. Then pick one and test it cheaply.

A worked example: five notes, one repeated problem

This is an illustrative set of made-up note excerpts.

  • "Spent Sunday chasing two clients for invoices again."
  • "Friend asked how I remind clients about payment without sounding rude."
  • "Idea: app for meal planning?"
  • "Third month in a row an invoice is 30 days late."
  • "Accountant says freelancers lose most time on admin, not taxes."

The repeated problem: late client payments for freelancers. It appears in three of the five notes, and in one of them it cost a whole Sunday.

The idea that stays an idea: "app for meal planning" appears once, with no person and no situation. Without a problem attached, it is a slogan.

A cheap test: ask five freelancers "When did a client last pay late, and what did you do about it?" If nobody can name a recent case, drop the idea. If most of them describe the same awkward reminder email, you found a real problem.

Practice a foreign language in a real situation

Textbook dialogues never go the way real ones do. Practice is useful when you choose the situation, try first, and only then see a correction.

Tell @vustbot the language, your level and the situation, and ask it to play the other person. Write your reply before you ask for any correction. This is an illustrative exchange in Spanish at a beginner level.

Situation: buying something for a sore throat at a pharmacy.

Pharmacist (played): "Buenos días. ¿En qué le puedo ayudar?"

Your attempt: "Quiero algo para mi garganta duele."

The smallest correction: "Me duele la garganta. ¿Tiene algo para el dolor de garganta?" In Spanish the throat "hurts to you", so the pain takes "me duele", and the medicine is "for the pain".

Next turn, with a new constraint: "¿Desde cuándo le duele?" Answer using "desde hace". A good reply is "Desde hace dos días."

Check: the next answer reuses "me duele" or "desde hace" without looking back. Using the correction in a new sentence is the sign it stuck.

Spot the ideas you keep repeating in your own writing

After a year of posts or newsletters, you start to suspect you write the same piece over and over. Sometimes that is a theme worth owning. Sometimes it is a rerun your readers have noticed.

Paste four or five pieces into @vustbot, label them T1 to T5, and ask for repeated ideas with a quote from each piece where the idea appears. This is an illustrative example.

Repeated idea: "small teams move faster than big ones."

  • T1: "Five people shipped what fifty could not."
  • T3: "Every person you add is a new meeting."
  • T4: "We hired slowly on purpose."

The editorial decision: T1 and T3 make the same argument about speed, so merge them into one stronger essay. T4 uses the idea in a different context, hiring, so keep it separate. A repeated phrase is not always a repeated piece.

Check: every quote can be found word for word in the piece it is attributed to.

Turn a grandparent's memories into a family story

The stories come out at family dinners, a little different each time, and nobody writes them down. A family history needs two things: the memories in the storyteller's own words, and a clear line between what they remember and what is confirmed.

Record the conversation and send the recording to Jot Bond (@vustScribeBot) for a transcript. Then paste the transcript into @vustbot and ask for a short chapter that quotes the storyteller and marks every name and date that is still uncertain. This is an illustrative excerpt.

From the transcript: "We came to the city the winter the river froze, I think 1956, maybe 1957. My father worked at the bread factory near the station."

In the chapter: "Grandma remembers arriving in the city the winter the river froze, in 1956 or 1957. Her father worked at a bread factory near the station."

Questions for the next visit:

  • Which year was it, and does anyone have a photo or document from that winter?
  • What was the name of the bread factory?
  • Who else came to the city with the family?

Check: the chapter never turns "I think 1956" into a fact. Read the chapter back to the storyteller and correct it together before you share it with the family.

Open @vustbot and paste the notes, writing or transcript you want to work with.

Turn your notes into a talk that fits the time

The slot is ten minutes, your notes would fill thirty, and the slides are still a list of things you know. Most talks run long for one reason: they are organized around what the speaker knows instead of the one question the audience came with.

Start from that question. Give @vustbot the question, who is listening and your notes. Ask for a slide outline with one message per slide, then speaker notes for each slide. Paste the notes into the speaking-time calculator to see how long they take to say.

A worked outline: one question, six slides

This is an illustrative outline for a made-up ten-minute talk to a leadership team. The numbers are placeholders for your own data.

The audience's question: "Should we move customer support from email to live chat?"

  1. The answer first. Yes, for order questions, which are 60 percent of tickets in your data.
  2. What customers wait for today. Your current email response time.
  3. What chat changes. Order questions answered while the customer is still on the site.
  4. What it costs. Staffing hours you supply from your own plan.
  5. The risk. Complex refunds still need email, so the team runs both.
  6. The decision we need. Approve a one-month pilot for order questions only.

Every slide answers one part of the leadership question. A slide that answers nothing the audience asked, such as the history of the support team, is the first one to cut.

Speaker notes you can actually speak from

Speaker notes are cues, not a script. Here is slide 5 as a cue card:

  • refunds stay on email
  • two queues for one month
  • I will report both numbers at the end of the pilot

Three short lines remind you of the point without tempting you to read aloud. Any number in your notes should come from your own data. If a figure is a guess, say so on the slide or leave it out.

Check the timing before the rehearsal

Paste your full speaker notes into the speaking-time calculator. At a normal speaking pace of about 130 words per minute, a ten-minute talk is roughly 1,300 words of spoken text. Then rehearse once with a timer, because pauses, questions and slide changes always add time. If the rehearsal runs long, cut a whole slide rather than speaking faster.

Open @vustbot, send the audience's question and your notes, and ask for the six-slide outline.

Pull the decisions out of a pile of notes

Three meetings, two phone calls and a shared doc later, the notes say everything and decide nothing. Reading them again takes an hour, and the one thing you need, what was actually agreed and what is still open, is spread across all of them.

Paste the notes you want to combine into @vustbot as one message, and label each one, N1, N2, N3. Ask for three lists: what was found, what was decided, and what is still open. Every line points back to the note it came from.

A worked synthesis of three notes

This is an illustrative set of made-up notes about choosing an office.

  • N1, call with the landlord: "Second floor available from May. Rent fixed for two years. Parking not included."
  • N2, team meeting: "Everyone prefers the second floor. Agreed to sign if parking is under $200 a month."
  • N3, email from facilities: "Nearest parking garage quotes $240 a month per space."

Found:

  • The second floor is available from May with rent fixed for two years. (N1)
  • Parking costs $240 a month per space nearby. (N3)

Decided:

  • Sign for the second floor, on the condition that parking is under $200 a month. (N2)

Still open:

  • The $240 quote is above the agreed limit, so the condition for signing is not met. Either the team accepts $240, or someone asks the landlord to include parking. (N2 and N3 together)

The open item is the useful part. No single note says it. It only appears when the condition from one note meets the number from another.

Check the synthesis before you act on it

  • Every line names its note. A finding with no source is a guess, so remove it or find where it came from.
  • Decisions are only things someone agreed. "Everyone prefers" is a preference. "Agreed to sign if" is a decision with a condition.
  • Only the notes you pasted count. If a phone call never made it into the notes, it is not in the synthesis, so add it before you decide.

Open @vustbot, paste the notes with their labels, and ask what is still open.

Practical use cases

What people use @vustbot chat for

Model switchers
Ask the same question to GPT-6.1 Sol, Claude Opus 5.5 and Gemini 3.8 Flash without leaving the conversation.
Compare the answers side by side and keep the best one — no tab-juggling, no four separate subscriptions.
No-card users
Reach frontier models from a country or device where Western card subscriptions don't work.
Start with a free first answer, then pay by card inside Telegram — no bank card, no extra apps, no website sign-up form.
Mobile-first people
Run AI from the same Telegram app that's already open all day, on phone or desktop.
Frontier answers in-chat, with the conversation history kept in the thread you can scroll back to.
Cost-watchers
Default to a fast model for everyday questions, escalate to a strong model only when a task is high-stakes.
A genuine free tier for simple asks, plus metered access to GPT, Claude, Gemini, Grok and more when it matters.
How it works01–04

How model-switching works

  1. 01

    Open @vustbot in Telegram

    No app install and no sign-up form — your existing Telegram identity is the login.

  2. 02

    Ask your question

    vustbot sends simple asks to a fast model automatically, or you pick a specific model yourself.

  3. 03

    Switch models mid-conversation

    Re-ask with a different model and the thread context carries over, so comparing answers is one tap — not a copy-paste across four browser tabs.

  4. 04

    Pay only when you go strong

    Fast everyday models are free. Frontier models draw from a daily and monthly allowance, or you pay by card inside Telegram.

Same tool · in Telegram@vustbot

Try it in Telegram

Open @vustbot and switch between frontier models in one chat — first result free, no card to start.

Open in Telegram
Quality & trust

What's actually in the lineup

Real frontier models, not a thin wrapper

The current catalog spans GPT, Claude, Gemini, Grok, Kimi, DeepSeek, GLM, Qwen, MiniMax, Inkling, Muse, LongCat and Fugu. A trust footer shows which model answered.

A free tier that actually works

Your first answer is free, with no credit card required to get started; after that, passes are bought inside Telegram.

One thread, many models

Unlike single-model apps, you switch providers inside one conversation and keep the context — so checking a second opinion is a re-ask, not a new chat in a new app.

access inside Telegram

Pay by card inside Telegram instead of a Western credit-card subscription.

FAQ

Frequently asked questions

Can I try the AI chat in my browser without signing in?

Yes. The public demo gives two successfully delivered vust Fast answers in one contextual thread. Failed attempts do not use that allowance. Telegram sign-in then preserves the same conversation; the guest copy expires after 24 hours if it is not claimed.

Which AI models can I use in @vustbot?

22 models in one chat: GPT, Claude, Gemini, Grok, Kimi, DeepSeek, GLM, Qwen, MiniMax, Inkling, Muse Spark, LongCat and Fugu. The model picker page lists every current version.

Can I switch models in the same conversation?

Yes — that's the point. Re-ask with a different model and the thread context carries over, so you can put the same question to GPT and Claude back to back and compare. It's one chat, not four separate apps.

Is there a free version, and do I need a credit card?

Your first answer is free, with no credit card to start. When you want more frontier access you pay by card inside Telegram, not a card subscription.

How is this different from ChatGPT or a single-model app?

ChatGPT gives you one model. Single-model apps lock you into one provider. @vustbot lets you switch between several frontier models inside one Telegram conversation, keep the context, and pay only when you escalate to a strong model — without four subscriptions or a row of browser tabs.

Does it work without a Western credit card or extra apps?

Access is through Telegram, and you pay by card inside Telegram rather than a card-based website subscription. In-app checkout availability and pricing are managed inside Telegram — check the current options in the bot. This is the main reason people in card-restricted regions use it.

Is Grok / DeepSeek / Gemini actually included?

Yes. Grok, DeepSeek, Gemini, GLM and Kimi are in the current catalog alongside GPT and Claude. The bot always shows which model actually produced each answer.

Ready when you are

One chat. Every model.

Stop juggling four AI subscriptions and a wall of browser tabs. Ask once, switch models, keep the best answer — inside the Telegram you already use.