Use models
Open models, no GPU needed.
- 1Install the Braise desktop app and sign in.
- 2In your AI tool, enter the local API address and key that Braise gives you.
- 3Your prompts run on community machines and are paid for with promotional credits.
Community inference network · Public beta in preparation
The inference network run by its community.
Braise runs your prompts on open models hosted by idle computers in the community. Use them without owning a GPU, or share your own machine and earn Compute Credits for each job it completes.
Free public beta in preparation, no download yet. Promotional credits are not money: they can’t be transferred or converted, and balances reset before paid V1.
# the Braise app shows your address and key
curl $BRAISE_API/v1/chat/completions \
-H "Authorization: Bearer $BRAISE_KEY" \
-d '{"model": "braise/qwen3.5:9b",
"messages": [{"role": "user",
"content": "Summarize this public changelog"}],
"stream": true}'- 00.04sSent to an available community machine
- 00.21sThe model runs on that machine
- 02.87sDone · paid in Compute Credits
Why Braise
Most powerful computers sit idle for most of the day. Braise pools them into one inference network. Anyone can use open models on it, and anyone who shares a machine gets compute back.
Our goal: community inference that’s as reliable as a cloud API, and open to everyone.
- 0
- GPU, runtime or local model needed to use the models
- 250,000
- promotional credits (CC) given once to each verified beta account
- 3
- runtimes supported for sharing: Ollama, oMLX and llama.cpp
- 2
- beta desktop platforms: macOS on Apple Silicon and Windows x64
How it works
Use it, share it, or both.
The two roles are independent. You don’t need to share a machine to use models, and you don’t need to use models to share one.
Share a machine
Put an idle machine to work.
- 1Run a supported model with Ollama, oMLX or llama.cpp.
- 2Turn on sharing in Braise. Nothing runs until you do, and when you stop, jobs in progress finish first.
- 3Earn Compute Credits for each completed job, and spend them on the network.
Anatomy of a request
What happens to your prompt.
Each step, and who can see what.
- 01
Your AI tool
A supported app sends the request to the Braise API on your computer.
- 02
Braise app
Checks your account and asks the network for an available machine.
- 03
Coordination server
Picks a machine that serves the model and has capacity. On the community path it only handles routing and billing data, never the content of your prompts.
- 04
Community machine
Runs the open model on its own hardware and streams the answer back.
The machine that runs your request can read it.
Encryption protects data in transit, not on the machine doing the work. Use public or made-up data, never anything confidential.
No hidden backup provider.
If no suitable machine is online, the request fails with a capacity error and you’re not charged. It is never passed to a commercial API instead.
Model catalog
Open models, hosted by the community.
The beta starts with a short list of Qwen models. Each card shows which runtimes and platforms can host it.
- 0.8B
Qwen 3.5 0.8B
braise/qwen3.5:0.8b256K context
- Runtimes
- Ollamallama.cppoMLX
- Host platforms
- Mac · Apple SiliconWindows · x64
- 9B
Qwen 3.5 9B
braise/qwen3.5:9b256K context
- Runtimes
- Ollamallama.cppoMLX
- Host platforms
- Mac · Apple Silicon
- 27B
Qwen 3.8 27B
braise/qwen3.8:27b256K context
- Runtimes
- Ollamallama.cppoMLX
- Host platforms
- Mac · Apple SiliconWindows · x64
Availability depends on how many suitable machines are online for each model. Every beta user can use every available model, whatever platform hosts it.
Commitments
What we commit to.
Clear about privacy
We tell you up front that the machine running your job can read it. Our demos and first tests only use non-sensitive data.
Credits, not cash
Providers earn Compute Credits to spend on Braise. During the beta, credits can’t be bought, transferred, withdrawn or converted.
Community machines only
Every job runs on a machine from the community. No hidden commercial backup, no made-up capacity.
Figures you can check
We publish network figures only once the sample is big enough to mean something. Nothing inflated, nothing hidden behind a zero.
Comparison
How Braise compares.
| What you do today | With Braise |
|---|---|
| Pay for a commercial API | Use supported community models through a local API. Compatibility is confirmed tool by tool. The beta runs on promotional credits, so it can’t tell you yet whether you’d save anything. |
| Run models on your own computer | Use community models without buying a GPU or managing a model yourself. You still install Braise and set up a supported tool. |
| Rent a GPU by the hour | You don’t reserve a GPU: credits pay for completed jobs. There’s no payment during the beta. |
| List your GPU on a marketplace | Share a supported model without writing a listing or setting a price. You earn temporary credits to use on the network, not income, and your machine’s running costs are yours. |
Beta opening
Get notified when the beta opens.
Tell us if you want to use models, share a machine, or both. We’ll email you when the free public beta opens. This doesn’t create an account or reserve a place.
The beta uses promotional Compute Credits, which are not money. They can’t be bought, transferred, cashed out or converted, and all beta balances reset before paid V1.
FAQ
Frequently asked questions
What is community inference?
Running AI models on computers shared by a community of people, rather than in one company’s data center. With Braise, your request goes to a community machine that runs the open model you asked for and sends back the answer.
Does Braise work with my tools?
Braise provides a local API with /v1/models, /v1/chat/completions and /v1/responses endpoints. We confirm compatibility tool by tool, so we don’t claim that every OpenAI-compatible app, editor or agent works.
Which platforms does the beta support?
The first beta builds are for macOS on Apple Silicon and Windows x64. Linux and Intel Macs aren’t supported yet.
Is the public beta open?
Not yet. You can sign up to be emailed when it opens. From then on, anyone can create an account with a verified email: no invitation code, no manual approval. Signing up for the email doesn’t create a Braise account.
What do providers get?
Promotional Compute Credits for each completed job. They can be spent on Braise during the beta, but can’t be bought, transferred, cashed out or converted. Providers pay their own electricity, bandwidth and hardware costs, and no volume of work is guaranteed.
Do beta credits carry over?
No. Every beta balance, granted or earned, goes back to zero before paid V1. It won’t be converted into a paid balance, money or cryptocurrency.
How can I try the beta as a user?
When it opens, each eligible verified account gets 250,000 promotional CC, once. It’s not a daily allowance, a cash amount or a set number of requests. Install Braise and set up a supported tool. You don’t need a GPU, a runtime or a local model.
Is Braise cheaper than my current API?
We can’t say yet. The beta is free and runs on promotional credits, so it can’t be compared with a paid price. Any future comparison will give real prices, comparable workloads, dates and measurements. We won’t promise a percentage of savings.
Can the machine running my prompt read it?
Yes. The community machine that runs your request can read it and its response. Encryption protects data in transit, not on the machine doing the work. Use public or made-up data, not confidential information.
What if no machine is available?
Your request fails with a capacity error, and you’re not charged for inference. It isn’t sent to a commercial API instead. Availability depends on how many suitable machines are online for the model you ask for.
What do I need to share my machine?
A platform, runtime and model combination listed in the provider guide, and enough free capacity to run it. Owning a GPU isn’t enough on its own. Sharing is optional: you can use Braise without ever providing.