AI Power Grid network
Worker software

Run AI Power Grid

Put a supported GPU to work serving decentralized text, image, video, and audio jobs. Start with the live text worker today; the media manager remains release-gated until its signed hardware profiles qualify.

Choose your operating system

Final compatibility is checked locally.

v0.3.5
Download text worker for Linux44.3 MB
Ubuntu 22.04+ x86_64Release notes

Live text worker · qualified media next

The text worker connects to your existing Ollama, vLLM, SGLang, LMDeploy, LM Studio, or KoboldCpp backend. The first managed media profile targets ACE-Step audio and stays unavailable until its signed qualification evidence is complete.

  • Text release live
  • WebSocket dispatch
  • Media fail-closed

Operator planner

Find the useful path for your machine

Enter coarse hardware details locally. Nothing is submitted. This planner narrows the worker path; only a signed profile and local canary can approve an exact capability.

Operating system

Likely starting point

Media candidate; use the live text path today

Your entered hardware clears the draft ACE-Step minimums, but the public media manager is still qualification-gated. This is a candidate result, not permission to serve yet.

You are also a strong candidate for the operator-managed ComfyUI image/video path; exact workflows and checkpoints still need a local test.

Entered accelerator
NVIDIA
Expected text speed
Not entered

Live network opportunity

Capacity and actual 30-day work

Jobs per worker is a rough workload-share signal, not a payout forecast. Throughput is observed across the existing Grid and is not a benchmark for your GPU.

Smollm-135m

text

Single-worker risk
Workers
1
Jobs, 30d
9,358
Jobs / worker
9358.0
Observed
375.6 tok/s

deepseek-v4-flash-nvfp4

text

Single-worker risk
Workers
1
Jobs, 30d
4,609
Jobs / worker
4609.0
Observed
91.1 tok/s

Krea 2 Turbo

image

Single-worker risk
Workers
1
Jobs, 30d
97
Jobs / worker
97.0
Observed
8.8s average

z-image-turbo

image

Single-worker risk
Workers
1
Jobs, 30d
61
Jobs / worker
61.0
Observed
23.4s average

LTX-2.3

video

Single-worker risk
Workers
1
Jobs, 30d
47
Jobs / worker
47.0
Observed
65.9s average

FLUX.2 Klein 4B FP8

image

Single-worker risk
Workers
1
Jobs, 30d
28
Jobs / worker
28.0
Observed
5.7s average

ace-step-v1.5-xl-turbo

audio

Single-worker risk
Workers
1
Jobs, 30d
26
Jobs / worker
26.0
Observed
14.3s average

LTX Director 2.0

video

Single-worker risk
Workers
1
Jobs, 30d
0
Jobs / worker
0.0
Observed
No recent timing sample

LTX-2.3 Audio

video

Single-worker risk
Workers
1
Jobs, 30d
0
Jobs / worker
0.0
Observed
No recent timing sample

gpt-oss-120b

text

One replica short
Workers
2
Jobs, 30d
4,099
Jobs / worker
2049.5
Observed
45.5 tok/s

Local decision

The browser does not decide what your GPU can run.

Private hardware check

The manager detects GPU, VRAM, driver, RAM, disk, and architecture locally. The Grid receives a capability tier and measured performance, not your complete hardware inventory.

Fail-closed profiles

Exact source, dependencies, model files, and recipes are verified before installation and again before serving. A capability stays unavailable until its local canary passes.

One path

From download to paid work

  1. 01

    Verify and install

    The manager chooses a supported signed profile, resumes downloads, and verifies every committed byte.

  2. 02

    Test and connect

    A real local generation proves the profile works. Then Google or wallet sign-in links the rig to a payout wallet delegation.

  3. 03

    Serve and earn

    The worker accepts compatible Grid jobs, signs output receipts, and earns through the public Base payout rail.