Skip to inference
Ephemerent / Colony

FREE COMMUNITY INFERENCE · PREVIEW

Small contributions.
Shared intelligence.

A place for spare GPU capacity to become something useful. Try a model for free when contributors are online—or help bring the next one online.

Try a prompt Bring your GPU

No paid fallback. Availability follows the community.

01 / THE PLAYGROUND

One prompt. Shared compute.

Free · capacity limited
0 / 2,000

One request at a time. Up to 2,000 characters and 6,000 UTF-8 bytes. No conversation history is sent.

A public experiment, not a private workspace.

Volunteer GPU operators can see your prompts and responses. Never send passwords, API keys, personal data, or confidential information. We cannot guarantee what volunteer operators retain.

Responses are capped at 256 tokens. Capacity and request limits may apply.

Response

NOT STARTED

Room for an idea.

Your model’s response will appear here.
Nothing is generated until you send a prompt.

Generated answers can be wrong. Check important details.No conversation saved in browser storage.

Checking whether volunteer capacity is available…

02 / A COMMUNITY RESOURCE

Many machines.
A common purpose.

Colony connects requests to available contributor capacity. When the network is full or offline, requests stop—not switch to a paid service.

  1. 01

    Contributors opt in

    People choose to offer GPU time. They control their machines and when they participate.

  2. 02

    Capacity becomes available

    The coordinator reports connected workers and models. The playground only enables requests when capacity is reported.

  3. 03

    You send a single prompt

    A volunteer-backed model returns a short response. Capacity can change mid-request; there is no uptime guarantee.

03 / BUILD IT WITH US

Your spare GPU.
Someone’s next idea.

Contributor onboarding is invite-only during the preview. Tell us your GPU model, VRAM, and when you’d like to help.

Request a contributor invitation

Always an explicit choice.

Opening this page does not install software or use your GPU. Participation requires a separate setup and your agreement.

Inference is free to users. Contributors cover their own electricity, hardware, and bandwidth. No payment or compensation is promised.