First training run starts this September — connect your GPU and be part of it.Contribute →

HADAL RESEARCH

CONTRIBUTE TO THE DEPTH.

Hadal Research is built by people contributing what they can.
Choose how you want to participate.

First training run starts this September. Set up your worker now — online machines get the first shards automatically. pip install hadal-worker

01

CONTRIBUTE COMPUTE

Run the Hadal Worker and contribute authorized GPU resources. Sandboxed jobs, your files untouched.

$pip install hadal-worker && hadal-worker
Run worker →
02

CONTRIBUTE DATA

Submit datasets for the shared shelf — annotations, preference rankings, domain knowledge. Reviewed before approval, then used by training runs.

03

EVALUATE MODELS

Independent evaluations from many machines are how results get verified. Eval shards are distributed to workers automatically with every run.

See runs →
04

CONTRIBUTE RESEARCH

Submit experiments, ideas, methods, and discoveries to active runs — or open an issue in any HadalAI repo.

GitHub ↗

SAFETY / Jobs execute sandboxed. The worker never reads personal files. Only hardware you own. Resource limits are yours to set.