Share by choice
Share Compute is opt in. Joining a conversation does not automatically contribute a device.
How Ameeba Compute works.
Share available Pseudopod computing power with the workspace you choose. Pool devices for throughput or cluster compatible hardware for a larger AI workload.
From workspace to compute
One switch connects available power.
Share Compute makes eligible Pseudopods available to that collaboration without turning every participant into a compute contributor.
WorkspacePeople and Max are working together.
→Share ComputeYou explicitly make available hardware eligible.
→PseudopodsCompatible devices contribute available GPU resources.
Two ways to computePool for throughput.
Pool for throughput.
Cluster for scale.
Compute Pool
More work at once.Independent Pseudopods can claim independent Max jobs. Adding workers increases cumulative throughput without pretending their VRAM is one GPU.
Compute Cluster
More power for one workload.Compatible nodes can cooperate on one model or inference workload. Model, runtime, network, and hardware compatibility matter.
What devices contribute
VRAM matters when AI models need it.
Ameeba surfaces contributed VRAM because model capacity is constrained by GPU memory. It is a capacity signal, not a promise that separate GPUs behave like one larger GPU.
8 GB VRAM12 GB VRAM16 GB VRAM
Control
Your hardware joins only when you choose.
Scoped to the workspace
Eligible compute follows the collaboration where sharing is enabled and access follows current membership.
Turn it off
Disabling sharing prevents new shared jobs from reaching your device. Work already claimed may finish.
Trust boundary
Remote compute processes the workload.
Encrypted transport protects data while it moves. It does not make a remote host unable to inspect the work it processes. Share compute with devices and participants you trust.
Open Security Center →