# Nerf Bench

> Grain re-runs a fixed task set against each model on a schedule to show whether a model gets worse over time, with per-model pass-rate trends and drop alerts.

- Canonical page: [/nerf-bench](https://grain.sh/nerf-bench)
- Agent-readable page: [/agents/nerf-bench.md](https://grain.sh/agents/nerf-bench.md)
- Section: Benchmarks

## What this page covers

- Fixed task set
- Scheduled re-runs
- Pass rate trend per model
- Regression alerts

## About Grain

Grain is a multi-agent work platform. A person describes an outcome as a Mission, Grain coordinates the models, tools, files, repositories, and connected computers needed to do the work, and the result returns for review.

Start with [the complete agent-readable index](https://grain.sh/agents/index.md) or read [the full site corpus](https://grain.sh/llms-full.txt).