Limonata
AI COMPUTE · TESTNET

The people powering Limonata AI.

Follow community compute from a connected GPU to accepted work. Find any public payout address and explore its contribution history.

Contributor addresses

2
Grouped by public payout wallet

Registered workers

4
Includes offline worker identities

Connected workers

1
Currently in contact with the coordinator

Ready workers

0
Ready for complete requests. A loaded shared stage may still be waiting for its partner.

SHARED MODEL · TESTNET

One Qwen3-32B model, across GPUs.

A connected GPU can still be downloading or loading. The complete model becomes available only when every required stage is ready.

Waiting for all model stages

1 / 2 required stages ready

0 active chats · 0 queued · — model slot capacity

Stage 1

Not ready

Layers 0–47

0 ready · 0 assigned workers

Initial admission: 18.2 GiB free VRAM, followed by a real GPU probe.

Stage 2

Resident and ready

Layers 48–63

1 ready · 1 assigned workers

Initial admission: 8.2 GiB free VRAM, followed by a real GPU probe.

Shared-model work is sponsored and currently earns no LIMO. Extra compatible workers can provide capacity or warm replicas. Waiting and unused replicas do not count as completed work.

Contributors

Current availability and retained history, one row per payout address.

0x88a225…2188e9View contribution

Qwen2.5-1.5B · Qwen3-32B · Qwen3-32B · shared model

Standby

1 / 3 connected

12 GiB VRAM
provider-reported

Historical response rate

—

No observed sample yet

Accepted participations

0

0 sponsored

Settled earnings

0 test LIMO

Cumulative · not a wallet balance

Offline

0 / 1 connected

— VRAM
provider-reported

Historical response rate

—

No observed sample yet

Accepted participations

0

0 sponsored

Settled earnings

0 test LIMO

Cumulative · not a wallet balance

Updated 2026-10-17 23:44:35 UTC · refreshes every 15 seconds

Contribute your GPU

Work and earnings recorded since 2026-10-17 22:50:24 UTC. Earlier worker-only records are excluded when their payout cannot be verified.

What these numbers mean

Useful work, recorded

Work counts provider participations in coordinator-accepted requests. Multiple GPUs can participate in the same shared request. These counts are not proof of work or independent hardware verification.

Response rate, in context

Characters/s uses coordinator time from claim to accepted answer, including preparation and network time. A whole-group rate describes the shared answer, not one GPU. Rates are historical, model-specific and are never added together.

Earnings follow settlement

Earned totals include settled paid work. Sponsored requests and idle time earn no coins. Escrow withdrawal balances can also include chat refunds and are separate from earnings. Test LIMO has no monetary value.