Models
Frontier and open models in one catalog. Chat, generate and switch models without opening new accounts.
See the catalogIntelligence for all
Every AI model. One private gateway.
Frontier and open models, an open-source hub you can run from, and a no-code agent studio, all behind one wallet, one credit balance and one API.
- Wallet sign-in
- Blind routing
- Powered by $ALL
Models
Hub
Agent Studio
Platform
Three surfaces, one foundation
Use models directly, run open-source work from the Hub, or build your own agents. All three share the same sign-in, balance and router.
- Phase 2
Hub
An open-source hub of models, adapters, agents and datasets. Request a deployment and run it on your credits.
Explore the Hub - Phase 2
Agent Studio
Build agents from blocks without writing code. Test them, publish them, and cap what they can spend.
Meet Agent Studio
- Private by design
- One credit balance
- One router
Modalities
Four modalities, one interface
Move between writing, images, video and voice without changing apps, accounts or payment plans.
Text
Chat, write, reason and code with leading language models.
- In
- Text, documents
- Out
- Text, code
Use cases
- Draft and edit long-form writing
- Debug and explain code
Image
Generate and edit images from a single prompt.
- In
- Text, images
- Out
- Images
Use cases
- Concept art and moodboards
- Product shots and mockups
Video
Turn a script or a still into a short video clip.
- In
- Text, images
- Out
- Video clips
Use cases
- Storyboards and animatics
- Social clips from a script
Speech
Convert text to natural speech, and speech back to text.
- In
- Text, audio
- Out
- Speech, transcripts
Use cases
- Voice-overs and narration
- Transcribe meetings and interviews
How it works
Connect. Stake or buy. Create.
- 01
Connect your wallet
No email, no password. Your wallet is your sign-in.
- 02
Stake or buy
Stake $ALL for a daily allowance, or buy a credit package.
- 03
Create
Use any model, Hub deployment or agent, all paid from one credit balance.
Privacy labels
Private or Blind: you choose
Every model, deployment and agent carries one of two labels, so you know what happens to a prompt before you send it.
- Private
Private
Runs on hardware allwit controls
- No outside model provider sees your prompt
- Covers allwit-hosted models and Hub deployments
- Prompts and replies are not stored
- Blind
Blind
Runs at an outside model provider
- The provider sees the prompt, never who sent it
- Your identity is removed before the request leaves allwit
- Opens up the frontier models
Labels appear in the catalog, in the model picker and on every Hub listing. An agent carries the label of its least private part: one Blind model or tool makes the whole agent Blind.
- 01
Anonymous tokens
Your credits become blind-signed tokens that can't be traced back to your wallet.
- 02
Sealed request, oblivious relay
The request is sealed on your device. The relay passes it on without being able to read it.
- 03
Gateway spends a token
The gateway opens the request, spends one token and routes it. It sees what, never who.
- 04
Nothing kept
Prompts and replies are not stored. Only token and usage counts remain.
Provider sees: an anonymous request
Anonymous tokens and the oblivious relay are in development. Today allwit removes your identity before a request reaches any provider.
Blind routing
Your prompts are nobody's business.
allwit splits knowing who you are from knowing what you asked, so no single party ever holds both.
No single party sees who and what
The relay knows where a request came from but can't read it. The gateway can read it but doesn't know who sent it.
Independent relays
Relays can be run by independent operators or $ALL stakers, so no one company holds both halves.
Sealed gateway on the roadmap
The gateway is planned to run inside a trusted execution environment (TEE), so even allwit's operators can't read requests.
Model catalog
One interface. Every frontier.
Pick the right model for the job. Each one shows its cost in credits up front.
- textFrontier
Claude
Anthropic
Careful reasoning, long documents and code, with a strong focus on safety.
- Privacy
- Blind
- Source
- Provider API
3 credits / use - textFrontier
GPT
OpenAI
General-purpose chat, writing and reasoning across a wide range of tasks.
- Privacy
- Blind
- Source
- Provider API
3 credits / use - textFrontier
Gemini
Google
Multimodal model with very long context for research and analysis.
- Privacy
- Blind
- Source
- Provider API
2 credits / use - textOpen weights
DeepSeek
DeepSeek
Open-weight model with strong reasoning and coding at a low cost per request.
- Privacy
- Blind
- Source
- Provider API
1 credit / use - textOpen weights
Qwen
Alibaba
Open-weight family with broad multilingual coverage and tool use.
- Privacy
- Blind
- Source
- Provider API
1 credit / use - imageFrontier
GPT Image
OpenAI
Image generation and editing that follows detailed prompts and renders text well.
- Privacy
- Blind
- Source
- Provider API
4 credits / use - imageFrontier
FLUX
Black Forest Labs
Photorealistic and stylised images with precise typography and layout control.
- Privacy
- Blind
- Source
- Provider API
4 credits / use - imageOpen weights
Qwen-Image
Alibaba
Open-weight image generation and editing with strong text rendering.
- Privacy
- Private
- Source
- allwit-hosted
2 credits / use - videoFrontier
Veo
Google
Cinematic text-to-video with native audio and high-resolution output.
- Privacy
- Blind
- Source
- Provider API
20 credits / use - videoFrontier
Kling
Kuaishou
Video with realistic motion and camera control, from text or a still image.
- Privacy
- Blind
- Source
- Provider API
15 credits / use - videoOpen weights
Wan
Alibaba
Open-weight text- and image-to-video for short clips.
- Privacy
- Private
- Source
- allwit-hosted
10 credits / use - speechFrontier
ElevenLabs
ElevenLabs
Natural, expressive text-to-speech voices in many languages.
- Privacy
- Blind
- Source
- Provider API
2 credits / use - speechFrontier
Deepgram Nova
Deepgram
Fast, accurate speech-to-text for recordings and live audio.
- Privacy
- Blind
- Source
- Provider API
1 credit / use - speechOpen weights
Whisper
OpenAI
Open-weight speech-to-text covering close to a hundred languages.
- Privacy
- Private
- Source
- allwit-hosted
1 credit / use - speechOpen weights
Kokoro
hexgrad
Lightweight open-weight text-to-speech with clear, natural voices.
- Privacy
- Private
- Source
- allwit-hosted
1 credit / use
Native allwit models join the same catalog later.
Credit costs shown are placeholders. Model line-up, privacy labels and availability may change. Model names are trademarks of their respective owners.
Hub
An open-source hub you can run from
Find community models, adapters, agents and datasets, then run them on allwit with the credits you already have.
What you'll find
Models
Open-weight models ready to deploy.
Adapters
Fine-tunes that sit on top of a base model.
Agents
Agents built in Agent Studio, ready to run or remix.
Datasets
Data for fine-tuning and for agent knowledge.
From listing to running
- 01
Publish
Creators publish a listing with its licence and a content hash.
- 02
Request a deployment
Anyone can ask for a listing to be served.
- 03
Serve
allwit runs it, and it joins the catalog with its privacy label.
- 04
Scale down
Deployments nobody is using scale down to free capacity.
Creators earn credits when others use what they publish. The share is set in the token model.
Agent Studio
Build agents without code
Snap blocks together to make an agent, test it, then publish it to the Hub.
Blocks
Models
Pick one or more models from the catalog.
Instructions
Tell the agent how to behave.
Knowledge
Give it documents to draw on.
Tools
Let it call tools and APIs.
Memory
Let it remember across runs.
Triggers
Run it on a schedule or an event.
- 01
Build
Arrange blocks and set instructions.
- 02
Test
Try it in a sandbox before anyone else can.
- 03
Publish
Share it on the Hub, or keep it to yourself.
Privacy
- Agent state is encrypted to your wallet. allwit stores only a reference to it.
- An agent carries the label of its least private block.
- Every run goes through the same blind router as any other request.
Spending caps
Set a credit cap per run and per day, so an agent never spends more than you allow.
Credits
One balance. Two ways to fill it.
Credits pay for every model, Hub deployment and agent run. Get them by staking $ALL, by buying a package, or both.
Stake $ALL
Receive a daily allowance sized by your stake. Your $ALL stays yours. The allowance expires at the daily reset.
Buy a package
Buy credits when you need more, or if you'd rather not stake. Purchased credits last longer than the daily allowance.
Credits that expire soonest are spent first.
Credit estimator
5,000 $ALL
100100,000
Estimated daily allowance
credits / day
Over 30 days: 75,000
Illustrative. Rates, thresholds and package prices are set in the token model.
Staking tiers
| Tier | Stake | Daily allowance | Suits |
|---|---|---|---|
| Starter | Entry stake | Base allowance | Trying models and light chat |
| Pro | Higher stake | Larger allowance | Daily work across models |
| Power | Highest stake | Largest allowance | Agents and heavy use |
Indicative only. Thresholds and allowances are set in the token model.
Infrastructure
Decentralised by design
allwit runs sign-in, staking and credits on a public blockchain, with $ALL at the centre.
Decentralised
Access runs on a public network, not on one company's account database.
Non-custodial
You hold your keys. Staked $ALL remains yours at all times.
Verifiable
Staking and credit rates are set on-chain, so anyone can check the rules.
Comparison
Stop juggling subscriptions
| Feature | Separate AI subscriptions | allwit |
|---|---|---|
| Accounts | A separate sign-up for every provider | One wallet sign-in |
| Billing | Several monthly subscriptions | A daily allowance from staking, or credit packages |
| Privacy | Prompts linked to your email and card | Blind routing removes your identity |
| Model choice | Limited to each provider's own models | Text, image, video and speech in one place |
| Switching cost | New account, new plan, new interface | Switch models in one click |
| Open-source models | Download and host them yourself | Run Hub listings on your credits |
| Agents | Separate tools, each with its own plan | Build and run agents in Agent Studio |
| Privacy labels | Data handling buried in the terms | Every model marked Private or Blind |
FAQ
Questions, answered
allwit is a private, decentralised AI aggregator with three surfaces: Models, for frontier and open models; the Hub, for running open-source work; and Agent Studio, for building agents without code. You sign in with a Web3 wallet and pay from one credit balance.
Roadmap
Where allwit is going
Phase 1Now
Models
Wallet sign-in, one credit balance and blind routing to frontier and open models.
Phase 2
Hub and Agent Studio
Run open-source work from the Hub and build agents without code.
Phase 3
Anonymous tokens
Blind-signed credit tokens and an oblivious relay, so no single party sees who and what.
Phase 4
Native models
allwit's own models join the catalog, behind a sealed gateway.
Your keys. Your models. Your privacy.
Connect a wallet to get started. No email, no password, no subscription.
Connect wallet