hipfire
/docs

documentation

Reference docs for hipfire — install, CLI, models, quantization, multi-GPU, benchmarks. Rendered from the warpfront/hipfire beta branch.

beta branch

Rendered documentation is pulled from warpfront/hipfire branch beta at build time. Edits land in the engine repo; this site republishes the curated set.

GPU diagrams live on /learn
Looking for how AMD GPUs compute?

Distilled decks on silicon math — not API reference. Start with RDNA / WMMA (consumer) or CDNA / MFMA (data-center), then the why-hipfire framing.

start here

Getting started

Install and first run.
CLI reference

Every subcommand.
Models

Curated tags & BYO.
Quantize

HF / safetensors / GGUF to hipfire formats.
Serve

OpenAI-compatible HTTP API.
Architecture

Engine layout, dispatch, and model paths.
Quantization design

MQ V2, HFQ, asym KV, and FWHT math.
Multi-GPU

PP / EP and memory budget.
Benchmarks

Measured performance per architecture.

also in the set