You have been using
one tenth of AI.
Most people meet artificial intelligence through a single chat box and quietly assume that is the whole thing. It is not. This page explains what is actually available, in plain words — no jargon you have to already know.
Six words, then everything else makes sense.
- Model
- One AI brain, built by one company. GPT is a model. Claude is a model. Gemini is a model. They are trained differently, so they are good at different things — and they charge different amounts.
- Skill
- Something a model can do beyond writing text: look at a picture, read a PDF, search the live web, think step by step before answering, or make an image. Not every model has every skill.
- Token
- How AI counts. Roughly three-quarters of a word. Your question and its answer are both measured in tokens, and tokens are what the model actually charges for.
- Context
- How much a model can hold in mind at once — the conversation so far, plus anything you attached. A bigger context means you can hand it a whole document instead of a paragraph.
- Credit
- Our single unit of spending, so you never have to track eight companies' prices in dollars. Your plan includes credits each month and you can top up in rupees.
- Consensus
- When several models have answered the same question, one model reads all of those answers and writes the single reply that reconciles them.
400+ models, and why that number matters.
There is no best AI. There is only the best AI for the thing you are doing right now. One model writes warm, natural prose and is expensive. Another is a tenth of the price and perfectly good at summarising an email. A third is slow but genuinely better at maths. A fourth understands Hindi the way a Hindi speaker does.
New Edge Soft Sol connects to over 400 models from more than 60 makers. You are not buying access to a company — you are buying access to the field. When a maker ships something new, it appears in the catalogue instead of becoming yet another subscription.
None of this asks anything of you technically. There is no key to generate, paste, store or rotate, and no bill from a foreign company in a foreign currency. You sign in and the models are there.
- OpenAI
- Anthropic
- Meta
- Mistral
- DeepSeek
- Qwen
- xAI
- Cohere
- Sarvam
- Perplexity
- Amazon
The catalogue is what the platform reaches. Which models your account can pick from depends on your plan — every plan's list is on the pricing page.
What a skill is, and why it changes what you can ask.
If you have only ever typed a question and read the answer, you have used one skill: text. Modern models do considerably more, and the difference between models is often not how clever they are but what they are allowed to touch.
Every model in our catalogue is tagged three ways — what it can read, what it can make, and what it can do. Those tags are not decoration. The composer reads them and adjusts itself: attach a PDF and models that cannot read documents step aside; turn on web search and the switch is only live where the model supports it; ask for a picture and only the models that generate images will answer.
This is why you never have to memorise which model does what. You choose what you want to do, and the product narrows the field for you.
Reads
What you can give the model- ImagesPhotograph a whiteboard, a receipt, a form or a rash and ask about it.
- DocumentsHand over a PDF, a contract or a spreadsheet and ask questions across all of it.
- AudioSpeak instead of typing, or pass in a recording.
- VideoSome models will watch a clip and describe or summarise it.
Makes
What comes back that is not text- ImagesDescribe a picture and it arrives in the conversation, as an answer you can vote on and re-run.
- AudioAnswers you can listen to, including in Indian languages.
- VideoWhere a model supports it, moving output rather than still.
Does
What it can do on its own while answering- Web searchLooks things up live, so an answer is not limited to when the model was trained, and cites what it found.
- ReasoningThinks in steps before replying. Slower and dearer, and markedly better on hard problems.
- ToolsCalls out to other systems mid-answer rather than guessing.
One question, six shapes of answer.
A chatbot gives you one box and one reply, which means you never learn what you were not told. Here the question stays constant and you choose the shape of the answer.
- When you do not know who to trust
Compare
Ask once and every model you selected answers at the same time, each in its own column. Disagreement between them is information — it is usually where the question was genuinely hard.
- When you already know what you want
Single model
One model, one column, the ordinary chat you are used to — except you can change which model it is at any point in the conversation.
- When you have to act on the answer
Consensus
Every column's answer is handed to one model, which reads them all and writes the single reply that reconciles them. One thing to read instead of four.
- When the first answer is not quite it
Vote & re-run
Mark the column that got it right, or run the question again — every version is kept and you can flip between them, each with its own cost.
- When you need a picture, not a paragraph
Image studio
Generated images live in the conversation as answers, so they can be voted on, re-run and revisited like any other reply.
- When someone else needs to see it
Share & fork
Send any conversation as a link. Whoever opens it can read the whole exchange, or fork it into their own account and carry on from there.
What it costs, and why it is one bill.
Subscribing to each major AI separately runs to a few thousand rupees a month, in dollars, on cards that foreign billing sometimes declines — and most people use a fraction of what they pay for.
New Edge Soft Sol works the other way round. Your plan includes an allowance of credits each month. Every answer spends credits according to what that model actually charged for the tokens it read and wrote — so a quick question off a cheap model costs almost nothing, and you only spend the premium rate when you deliberately choose a premium model.
It is all in rupees, topped up over UPI, and every single answer carries its own receipt: which model, how many tokens, how many credits. Nothing about the spending is hidden from you.
- Currency
- Indian rupees, end to end
- Top-ups
- UPI, cards and netbanking
- Receipts
- Per answer — model, tokens, credits
- To begin
- A free tier, no card
One chatbot vs. all of them.
| A single AI subscription | New Edge Soft Sol | |
|---|---|---|
| Models you can use | One company's | 400+ from 60+ makers |
| If a better model ships | Wait, or subscribe again | It appears in the catalogue |
| Seeing a second opinion | Open another tab, pay again | Side by side, one question |
| Settling disagreements | You read and judge | Consensus writes one answer |
| Setup | An account per company | One sign-in, no API keys |
| Paying | Per company, in dollars | One wallet, in rupees over UPI |
| Knowing what you spent | A flat charge either way | A receipt on every answer |
| Indian languages | Translated, if supported | 22 languages, native voice |
Honest about what is not built.
These are planned, not available. We would rather say so here than let you find out after paying.
Bring your own key
Coming soonPlug in your own provider API key and route your usage through it. Today every model is served through our own connections — there is no place to add a key, and no billing path for one.
Fair questions.
Do I need to know anything technical to use this?
No. There is nothing to install, no key to create, and no setting you must get right before it works. If you can use a chat app, you can use this — the model choices have sensible defaults and you can ignore them entirely.
Why would I want more than one model to answer?
Because a single answer never tells you how confident to be in it. When four models agree, you can move on. When they disagree, you have just discovered that the question was harder than it looked — which is exactly the moment you would have been misled by one confident reply.
Does using several models cost four times as much?
Comparing does cost more than asking one model, because four models genuinely do the work. But you control which and how many, cheaper models cost a small fraction of premium ones, and every answer shows you what it spent — so the cost is a choice you make rather than a surprise.
Can I use my own OpenAI or Anthropic API key?
Not today. It is on the roadmap, but nothing in the product currently accepts a personal provider key. Every model you use is served through our own connections, which is also why you do not need a key in the first place.
Do I get all 400+ models?
The catalogue is what the platform reaches. Which of those models your account can select is set by your plan, and each plan's list is published on the pricing page — so you can see exactly what you are getting before you pay.
Is my data used to train these models?
Your conversations are yours. What we retain, for how long, and what we never do with it is set out in full in our privacy policy.
Try it on a real question.
The free tier is enough to see all of this working. No card, no key, no setup.
Start free