> For the complete documentation index, see [llms.txt](https://docs.usedecentral.org/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.usedecentral.org/providers/browser-worker.md). # Browser Worker The browser worker is the zero-friction path to earning on Decentral. There is nothing to install. Open the Earn tab, connect a wallet, and start accepting inference jobs from your browser. Browser workers use WebGPU via the WebLLM runtime to run inference directly in the browser tab. Jobs are processed locally on your GPU and streamed back to the requesting client. *** ## Requirements | Requirement | Details | | ----------- | ---------------------------------------------------------------------------------------- | | Browser | Chrome 113+ or any Chromium-based browser with WebGPU enabled | | GPU | Any GPU with WebGPU driver support and at least 3GB of GPU memory visible to the browser | | Wallet | Any Solana wallet (Phantom, Backpack, Solflare) | | Network | Stable internet connection. Upload bandwidth affects streaming quality. | | Stake | Not required for browser workers | **WebGPU support check:** Navigate to `chrome://gpu` in Chrome and look for "WebGPU" in the graphics feature status. If it shows "Hardware accelerated," you are compatible. **Unsupported:** Safari (WebGPU implementation is incomplete), Firefox (WebGPU not enabled by default in most versions). Mobile browsers are not supported for worker mode. *** ## Getting started **1. Connect your wallet** Go to `usedecentral.org/app` and click "Earn" in the navigation. Connect your Solana wallet when prompted. This wallet address is where your USDC earnings will be sent. **2. Check compatibility** The Earn tab runs a compatibility check automatically when you connect. It detects: * Whether your browser supports WebGPU * How much GPU memory is available to the browser * Which models are compatible with your available VRAM If WebGPU is not available, you will see an error with a link to enable it or a suggestion to switch browsers. **3. Select a model** Choose from the list of compatible models. The list is filtered to models that fit in your detected VRAM. Selecting a larger model generally earns more per job but may slow other activity on your machine while jobs are running. | Model | Tier | VRAM required | Earnings per job | | ----------------- | -------- | ------------- | ---------------- | | Qwen3 1.7B (Q4) | Lite | 3GB | $0.015 | | Llama 3.2 3B (Q4) | Lite | 4GB | $0.015 | | Qwen3 8B (Q4) | Standard | 6GB | $0.060 | | Mistral 7B (Q4) | Standard | 6GB | $0.060 | **4. Start earning** Click "Start Earning." The page loads the selected model into browser memory. This takes between 15 seconds and 2 minutes depending on the model size and your internet connection. Models are cached in browser storage after the first load. Once the model is loaded, your worker announces itself to the orchestrator mesh and begins receiving jobs. You do not need to do anything else. The page shows incoming jobs and a live USDC earnings counter as jobs complete. *** ## How browser inference works When a job arrives at your browser worker: 1. The encrypted job payload is received via WebSocket from the orchestrator. 2. A Web Worker thread decrypts the payload using the ephemeral session key. 3. WebLLM runs inference on the decrypted prompt using your GPU via the WebGPU API. 4. Generated tokens are streamed back to the orchestrator as they are produced. 5. After the final token, the worker computes a SHA-256 hash of the full output, signs it with the wallet keypair, and submits the proof to the `settlement` program. 6. The settlement transaction sends USDC to your wallet. This typically confirms within one second. Inference runs in a Web Worker thread to keep the browser UI responsive during jobs. You can use other tabs while earning, though GPU-intensive tasks in other tabs may reduce inference throughput. *** ## Earnings and payouts Earnings are paid in USDC directly to your connected wallet after each completed job. There is no minimum payout threshold and no batching. Browser workers receive 75% of the job value. There is no stake requirement to start, but you cannot unlock the 85% rate with browser workers alone. The 85% rate requires the `decentral-node` native worker with an active $DECENTRAL stake. During the beta period, all browser workers receive a 2x $DECENTRAL reward multiplier on top of USDC earnings. This multiplier applies to $DECENTRAL allocation from the Community/Provider rewards pool and vests over 18 months. *** ## Leaving and returning Closing the Earn tab stops your worker immediately. Any job currently in progress will time out and be rerouted to another available worker. You will not receive payment for incomplete jobs at the moment of closure. Your USDC balance and on-chain reputation persist between sessions. When you reopen the Earn tab and reconnect your wallet, your history is intact. *** ## Troubleshooting **"WebGPU not available"** Enable hardware acceleration in Chrome settings (`chrome://settings/system`). If it is already enabled, check that your GPU driver is up to date. **Model loading fails** Your browser's available GPU memory may be less than the model requires. Try selecting a smaller model. Closing other tabs that use GPU resources (video players, 3D apps, other WebGPU applications) can help. **Jobs not arriving** After loading a model, it can take up to 60 seconds for your worker to be visible in the orchestrator mesh. If no jobs arrive after several minutes, check your network connection. Verify on the dashboard that your worker is showing as "Active." **Low earnings per hour** Browser workers on Lite-tier models earn less per job than native workers on larger models. If you have NVIDIA or Apple Silicon hardware, the \[Native Worker]\( ) path will earn significantly more per hour. --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter: ``` GET https://docs.usedecentral.org/providers/browser-worker.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.