GPU.md

# ABOUT GPU.MD

GPU.md is a working idea: a plain markdown file that tells an AI agent what one GPU can hold and how it may be used. Memory comes first: weights, context and batch all have to fit. A short file keeps an agent from loading a model that won't fit, or holding a card it isn't using.

## WHAT IT COULD COVER

• - the card: memory, supported precisions, power limit
• - what may run on it: models, sizes, precisions
• - who else shares it, and when
• - idle rules: let it go after a few minutes unused
• - what to do when memory runs out: smaller batch, lower precision, or ask

## A SKETCH

```md
---
gpu: draft
---
# GPU.md
- memory: 24 GB
- fits: 8B model, 8-bit weights, 16k context
- shared_with: nightly evals, 01:00-04:00
- release_after_idle: 15 min
- out_of_memory: halve batch, then ask
```

## DISCLAIMER

No prices, benchmarks or vendor claims are given or implied. Check fit with your own runtime before relying on it.

GPU.md is an informational page. It is provided as is, without warranty of any kind, and may be incomplete, inaccurate or out of date. Nothing here is financial, investment, legal, tax or other professional advice, or an offer or solicitation. The sketch is an illustration, not a standard or a recommendation. Names of third-party products or companies, if any, are used only to describe them; no affiliation or endorsement is implied. Use at your own risk. This page sets no cookies and runs no scripts or analytics; like any website, its host keeps standard request logs.