STEP 01
Register and join a pool
Sign up with your email. We assign a compute pool based on device class and region.
Mainnet v2.4 · 38,412 devices online
Your idle devices, now a priced compute asset
Rent out the spare compute in your phone, laptop or GPU. The client picks up matrix units from the cloud, runs them, submits the result — and you are paid in PEARL for work actually delivered. No staking, no power bill upfront.
Email and verification code only · free to start
38,412Devices online
131,368 PEARLSettled in 24h
3.42 PEARLAverage per device/day
96Regions
Onboarding
Four steps, then your device starts earning
From sign-up to the first settlement takes about 11 minutes.
STEP 02
Install the compute client
A lightweight process runs in the background, picking up work while the screen is off or the device is idle. Peak usage stays under 15%.
STEP 03
Receive matrix units
The cloud slices inference work into matrix shards and dispatches them by device capability.
STEP 04
Submit and get paid in PEARL
Settlement every 6 hours. Contribution details and on-chain hashes are always visible.
Architecture
Phones can run a large model, because they don’t run the whole model
Our inference fleet is built on DeepSeek V4.1 Flash. Only a subset of MoE experts is carved out and deployed to edge devices; aggregation happens in the cloud.
MoE expert sharding
Cold experts move to the edge by routing popularity; hot experts stay in the cloud GPU pool.
Cloud all-reduce
Shard outputs are reduced per layer in the cloud, with error below 1e-5.
Elastic scheduling
Offline devices never block a job — shards are rescheduled within 200ms.
Each token activates just a few experts, and cross-device traffic stays around 16Mb per token. Devices perform GEMM shards only — never a full forward pass.
Economics
Paid by contribution, not by luck
Weighted by effective FLOPs and uptime. PEARL lands in your balance every 6 hours.
| Device | Reference spec | Daily (PEARL) |
|---|---|---|
| Phone · tablet | 4–8 cores · 4GB+ | 1.2 – 2.8 |
| Laptop · desktop | Integrated GPU · 8GB+ | 4.5 – 9.6 |
| Discrete GPU | 8GB VRAM or more | 8.0 – 18.5 |
| Rig · server | Multi-GPU · 24h uptime | 22.0 – 46.0 |
Based on the median 3.42 PEARL per device per day. Actual payout follows measured contribution.
Estimate your payout
Devices
Online hours / day
Estimated daily
7.70PEARL
Based on the median 3.42 PEARL per device per day. Actual payout follows measured contribution.
Settlement asset
PEARL is the unit of account for this network
Total supply 2,100,000,000 PEARL, emitted entirely for compute work. No pre-mine, no private sale.
Total supply2.1B
Circulating684M
Settlement cycle6 hours
Withdrawal minimum50 PEARL
684,000,000 / 2,100,000,00032.6%
TRC-20 contract
TQm3xPEARL8sK4vN2pR7yLd9WcZ6hJf5Ub
Withdraw to any TRC-20 address. The network fee is covered by the protocol.
Questions
Before you plug in
By default the client only takes work while the screen is off or the device is idle. If it detects foreground load, it yields compute immediately.
About 3–7W extra per device. At residential rates, electricity is roughly one sixth of the daily payout.
No. Registering, installing and running are free. You are paid only after work is delivered.
Effective FLOPs multiplied by an uptime coefficient, distributed from the pool total every 6 hours.
Android 8+, Windows 10+, macOS 12+, plus mainstream NVIDIA and AMD GPUs.
On-chain within 10 minutes of requesting; a TRC-20 transfer usually confirms in under a minute.
PEARL NETWORK
Plug in before the next settlement round
38,412 devices are already running, averaging 3.42 PEARL per device per day.
Email and verification code only · no payment method required