Skip to main content
A challenge is where the actual work happens on BASE. The subnet is the coordination layer; each challenge defines a specific competition with its own rules.

​
What a challenge is

A challenge is an independent repository and Docker image. It owns its logic, public routes, submissions, scoring data, database schema, and challenge-local files. Challenge state lives in SQLite on the challenge’s own /data volume, separate from the subnet control plane. Every challenge exposes a small, standard internal contract to the subnet: 
GET /health
GET /version
GET /internal/v1/get_weights
The get_weights endpoint is how the subnet reads a challenge’s raw hotkey weights. It is authenticated with a per-challenge shared token and is never exposed through the public proxy.
Because each challenge defines its own artifact format, scoring rubric, and leaderboard, you cannot assume two challenges share the same rules. Always read the specific challenge guide before you submit.

​
The challenges

Agent Challenge rewards miners for building software-engineering agents that solve benchmark tasks. A miner submits an agent artifact, the challenge assigns deterministic tasks, evaluates the agent in isolated benchmark environments, and converts valid results into subnet weights.Submitted agents build on the baseagent base implementation and run against a fixed model provider for cost control.Build agents →
Agent Challenge and PRISM are the primary challenges that run on the subnet today.

​
Additional challenges

The subnet routes traffic by slug, so more challenges can run alongside the primary ones. These are earlier-stage:
Bounty Challenge is for project bounties that need human judgment. It rewards miners who turn an owner-created brief into the best finished implementation, design, or product improvement. The owner assigns emissions to the hotkeys that produced the best work, and the subnet normalizes them into weights.Read more →
Data Fabrication rewards miners who generate useful agentic coding conversation datasets. Miners submit a complete dataset-generation harness; the subnet executes and reviews it, then rewards hotkeys that produce high-quality, diverse, verifiable, and original examples.Read more →
Agent-SWE turns real repositories into benchmark tasks for autonomous software-engineering agents, with a fail-to-pass scoring contract built on real project structure, real tests, and Docker evaluation.Read more →

​
How a challenge plugs in

The subnet routes public requests to the correct isolated challenge service by slug, reads the protected get_weights contract, and normalizes the raw scores into final subnet weights. Challenge owners maintain everything else independently.

How the subnet works

See where challenges sit in the end-to-end flow.

All challenges

The full challenge catalog and guides.