hipfire

Local LLM inference for AMD GPUs. Tuned first for RDNA, with CDNA support.
Rust + HIP + Redline. No Python in the hot path. Pull, serve, chat.

Quickstart · What's new · Models · GPU support · PeaceMaker · Install · Docs

What hipfire is

hipfire is an inference engine for AMD GPUs. It ships its own Rust runtime, model implementations, quantization formats (MQ3–MQ6, MQ4R, MFP4), typed kernel dispatch and HIP kernels. ROCm is loaded dynamically. There is no Python, PyTorch, CUDA translation layer or third-party inference engine in the hot path.

You get a curated model registry, a CLI (pull, run, chat, serve, bench, quantize) and an OpenAI-compatible HTTP daemon. Speculative decoding (DFlash draft models, native MTP heads), multi-GPU expert parallelism and the Redline replay path ship with the engine.

Quickstart

hipfire pull qwen3.8:27b-mq4-xts
hipfire serve qwen3.8:27b-mq4-xts -d
hipfire chat qwen3.8:27b-mq4-xts

This downloads about 15 GB and needs a 24 GB+ card or a Strix Halo system. On smaller cards, use qwen3.5:4b or qwen3.5:9b with the same commands. Pick a model for your GPU lists recommendations for each GPU tier.

For one-shot inference:

hipfire run qwen3.8:27b-mq4-xts "What is the capital of France?"

Qwen3.8-Flash (qwen3.8:flash-next) is a MoE model. The tag resolves to the split base qwen3.8-flash-next-base.mq4-xts (about 70.9 GB), plus a shared per-token-embedding (PLE) sidecar (29.1 GB for q4 rows, 54.4 GB for q8 rows) and, on gfx1201, a 213 KB expert hot-set file. The old 125 GB monolith (qwen3.8-flash-next.mq4-xts, formerly qwen3.8-flash-next-gptq3.mq4) stays published for existing clients:

hipfire pull qwen3.8:flash-next
hipfire serve qwen3.8:flash-next -d

A 32 GB card alone is not enough for Qwen3.8-Flash. It needs a 128 GB Strix Halo system, or an R9700 with enough host RAM for the mapped experts, or two R9700s (EP2). Native context is 262K tokens, subject to available memory.

Image generation (RDNA3/3.5 measured):

hipfire pull flux.schnell:1
hipfire img --out x.png flux.schnell:1 "a red cube on a wooden table"

The daemon serves an OpenAI-compatible API on 127.0.0.1:11435. It listens on loopback only unless you set serve.host.

curl -s http://127.0.0.1:11435/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{"model": "qwen3.8:27b-mq4-xts",
       "messages": [{"role": "user", "content": "What is 2+2?"}],
       "max_tokens": 256,
       "chat_template_kwargs": {"enable_thinking": false}}'

Qwen models think by default, and max_tokens counts the <think> span as well as the answer. If thinking is on and the cap is small, content can come back empty with finish_reason: "length". Turn thinking off as shown above, or raise max_tokens. Qwen3.8 and Ornith 1.5 default to medium reasoning effort. See docs/SERVE.md.

What's new in 0.4.3

Every figure in this section comes from the v0.4.3 release notes. Each is a measurement on one named fixture and card, not a general throughput claim. The CHANGELOG 0.4.3 entry has the fixtures, methods and kill switches.

A3B prefill on the R9700: 0.4.2 (mq4r) vs 0.4.3 (mq4rs)

Qwen3.6-35B-A3B prefill on a Radeon AI PRO R9700 (gfx1201):

Prefill tok/s	0.4.2 (mq4r)	0.4.3 (mq4r)	0.4.3 (mq4rs)	Speedup
pp512	3,292	3,332	11,589	3.5x
pp1024	3,248	3,607	15,718	4.8x
pp4096	2,886	3,737	19,828	6.9x
pp8192	2,488	3,615	20,035	8.1x
pp16384	1,957	3,561	18,545	9.5x

The speedup column is 0.4.3 mq4rs over 0.4.2 mq4r. Most of the gain comes from the new qwen3.6:35b-a3b-mq4rs quant (18.7 GB): 3.5x to 5.5x faster than the same 0.4.3 engine on the old mq4r file. The 0.4.3 engine alone is 1.0x to 1.8x faster on mq4r, and it removes 0.4.2's long-context falloff (0.4.2 loses 41% from pp512 to pp16384; 0.4.3 on mq4r does not fall off).

Fast mq4rs prefill is gfx1201 only. On other GPUs the mq4rs file needs HIPFIRE_MOE_SYM_IU4=0, which makes prefill about 5x slower.

If you run the old mq4r, pull the new file:

hipfire pull qwen3.6:35b-a3b-mq4rs

Caveats:

• The measurement used one R9700, hipfire bench --matrix --pp 512,1024,4096,8192,16384 --tg 1 --ctx 1 --runs 3 --warmups 1 --spec off. Each cell is the median of 9 samples from 3 fresh processes, with the arms interleaved in one timing hold.
• "0.4.2" means v0.4.2.1. Each version ran its default config, and the defaults differ: 0.4.2.1 uses KV q8 and a 512-row prefill chunk, 0.4.3 uses fp8 KV and an 8192-row chunk, and the symmetric IU4 MoE prefill route is only admitted on the mq4rs file. The mq4r columns show what you get from upgrading without changing the model file, not a single-variable engine A/B.
• mq4rs WikiText-2 KLD is 0.129, against 0.206 for qwen3.6:35b-a3b-mq4r (same harness). There is no MTP sidecar for mq4rs yet, and its router and lm_head are 4-bit.
• On the old mq4r file, 0.4.3 prefill numerics change slightly (WikiText-2 prefill KLD 0.2082 to 0.2115, inside the quality bar), and greedy output can differ from 0.4.2 at near-ties.

Flash-Next on gfx1201 (R9700)

• Larger prefill chunk by default: faster prefill, slower decode. The default prefill chunk for Flash-Next is now 16,384 rows on gfx1201, together with the compact forward scratch (below). Against the same build at a 4096-row chunk (one R9700, gptq3 monolith, 3 fresh processes per arm), pp16384 goes from 5,405 to 6,431 tok/s (+19.0%). Decode pays for it: AR tg128 goes from 34.88 to 33.11 (−5.1%; synthetic tg128 uses random token ids and understates real-text decode), --spec auto prose from 49.1 to 44.3 (−9.8%) and code from 73.7 to 66.1 (−10.3%), and resident experts drop from 8,234 to 6,578. For short prompts with long generations, set HIPFIRE_PREFILL_CHUNK_ROWS=4096. Output is byte-identical.
• Compact forward scratch by default for symmetric packed Flash-Next (the split base and the gptq3 monolith). Measured at the 4096-row chunk on the split base: forward scratch 4,388 to 1,270 MiB (about 4.3 to 1.2 GiB), 8,234 experts resident instead of 6,948, pp16384 5,024 to 5,405 tok/s (+7.6%), tg128 33.56 to 34.84 (+3.8%). HIPFIRE_QWEN4_COMPACT_SCRATCH=0 restores the full scratch, and with it the 4096-row default chunk.
• Expert hot set by default. hipfire pull qwen3.8:flash-next fetches the 213 KB .expert-hotset sidecar and the load applies it on gfx1201 only. Hot experts stay resident in VRAM instead of whole layers. On real text it measured +14% to +26% MTP decode (fiction prose +14.4%, coding prompt +26.3%) and +7% to +40% plain AR, with no prefill loss (pp16384 +0.5%) and byte-identical output. The copy-heavy code edit under MTP is the exception (−3.1%). An earlier single-R9700 measurement, taken before the 16,384-row default and not re-measured since, found the hot set costs about 9.5% of short-context decode and 6% of pp512 against hot set off. HIPFIRE_QWEN4_EXPERT_HOTSET=0 restores whole-layer placement. Dense models and Strix Halo are not affected.
• Exact prefill kernels on by default on gfx1201: MQ6 INT8 GEMM (+3.8% at pp16384 on one R9700), plus four byte-identical prefill waves (W1–W4) whose gains are measured at the 16,384-row chunk. Kill switches are in the CHANGELOG.

Models, containers and tools

• LFM2.5 support. Decode runs as typed dispatch steps for LFM2.5 dense and MoE models, with correct think-channel and tool-call shapes, real stop tokens, and a fix for the MoE down kernel that made LFM2.5-8B-A1B emit degenerate text. LFM2.5-24B-A2B comes up with partial-GPU offload but has no registry tag in this release. The lfm2.5: tags are unchanged.
• Smaller Flash-Next tiers, qwen3.8:flash-next-mq3-xts and -mq2-xts; see the tier table.
• Container image. compose.yml runs a ROCm 10.1 full image (default) or a slim runtime image with baked gfx1100, gfx1151 and gfx1201 kernel packs. The image is published as ghcr.io/warpfront/hipfire:0.4.3.1 (and :latest).
• Strix Halo repetitive-prompt NaN fix, with no prefill cost. A long prompt with many repeated tokens could turn Flash-Next/Qwen GDN prefill NaN on Strix Halo and the 7900 XTX.
• New commands: hipfire fit, hipfire doctor, hipfire calibrate, hipfire doctor --npu and hipfire config tuning presets.
• Retired tags: qwen3.8:flash-next-rtn-asym and qwen3.8:flash-next-mq6q8-pleq8 are retired from new pulls; use qwen3.8:flash-next. Already-installed verified files keep running.

Upgrade notes and known issues are in the CHANGELOG.

What's new in 0.4.2

The chart compares the 0.4.1 Strix Halo release-matrix cells published on hipfire.dev (measured 2026-10-07) with the same lane rerun on v0.4.2 (2026-10-14): same artifact, host and method. The 16K NPU bar is new in 0.4.2 and comes from the release measurement of the shipped tip. Every figure in the list below comes from the CHANGELOG 0.4.2 entry that introduced the feature, which gives the full fixture, the method and the kill switch. Each figure is a measurement on one fixture, not a general throughput claim.

The sampled-MTP row is not reproducible token for token: with seed 123, 0.4.2 gives different token streams across fresh processes on three of the eight prompts (a known issue in the wall-time route chooser). Greedy AR and greedy MTP are bit-identical across processes.

• Qwen3.8-Flash expert parallelism (EP2 through EP8). Owner-routed prefill is on by default when topology.ep ranks of one exact arch, gfx1201 or gfx1100, admit it; long-prefill expert staging and balanced per-call expert ownership are gfx1201-only. Native MTP also runs under EP. EP is opt-in and experimental. Measured on 2x R9700 (gfx1201), pinned Qwen3.8-Flash XTS, committed prompts, 2026-10-13: prefill went from 3,033 to 3,943 tok/s at 8K (1.30x) and from 3,754 to 4,308 tok/s at 16K (1.15x), one card vs EP2. See multi-GPU serving.

• Qwen3.8-Flash NPU prefill route on Strix Halo. fusions.npu_trunk = "auto" runs the pipelined trunk input projections on the XDNA2 NPU. It needs exact gfx1151, an open NPU and a matching calibration. Any other system runs GPU-only. HIPFIRE_QWEN4_NPU_TRUNK=off disables it. Measured on gfx1151 + XDNA2, pinned Qwen3.8-Flash XTS, 16K-token prompt, 2026-10-13: TTFT went from 7,878 ms to 7,085 ms (−10.1%).

• Engine-level YaRN. The engine now reads YaRN RoPE scaling from the checkpoint. attention.yarn_factor / HIPFIRE_YARN_FACTOR overrides it, which lets Qwen3.8-Flash run past its native 262,144-token context. Measured on gfx1151, Qwen3.8-Flash GPTQ3 MQ4, yarn_factor = 4, 2026-10-13: a needle placed in a 499,207-token prompt was retrieved at 10% and at 90% depth.

• Halo sidecars (zab and PLE). On gfx1151, hipfire pull fetches the Qwen3.8-Flash zab trunk-A4 overlay, verifies it and applies it by default. The q4 PLE sidecar is also on by default there when present. hipfire ple rebind reuses a q4 PLE payload after the base is repacked. See configuration.

• Strix Halo 27B DFlash. On exact gfx1151, the 16-row verify GEMMs stage activations in LDS (HIPFIRE_GFX1151_XLDS, on by default). Measured on Strix Halo (gfx1151), Qwen3.8-27B XTS with the paired DFlash draft, committed code prompt, 2026-10-13: decode went from 97.9 to 135.3 tok/s (+38%). Output is byte-identical.

• Qwen3.8-27B DFlash-default prefill. When the registry-paired draft is installed, DFlash is the default, and cold prefill now keeps target hidden rows on the device. Measured on one R9700 (gfx1201), Qwen3.8-27B XTS with the paired draft, committed 8.2K-token prompt, 2026-10-12: prompt throughput went from 3,809 to 4,921 tok/s (+29%). See the paired DFlash default.

• MTP draft cap 4 on Halo. speculation.mtp_k = auto resolved to 4 for native Qwen3.8-Flash MTP on gfx1151 and to 3 everywhere else in 0.4.2. Measured on gfx1151, pinned Qwen3.8-Flash XTS, code prompts (Rust/Go/Java), 2026-10-12: geomean decode went from 67.9 to 70.5 tok/s (+3.8%). Since 0.4.3 auto drafts up to 7 on gfx1151 and the adaptive depth chooser picks each window's depth; see the CHANGELOG.

• Also in 0.4.2:

  • a faster Qwen35 MTP prompt fill
  • gfx1201 batched MTP fill for Qwen3.8-Flash
  • device-side sampled DFlash acceptance
  • continuous-batching sampled-lane batching on gfx1201
  • certified PeaceMaker kernels for CPU-AR sampling
  • symmetric-grid DFlash draft conversion
  • Windows iGPU detection
  • serve queue defaults taken from the config table

  See CHANGELOG.md for the full list.

BENCHMARKS.md explains fixture identity and the requirements for admitting a measurement. Its older tables are historical, not current product baselines.

Models

The bundled registry (registry/v1.json) contains 94 curated entries (models, quant variants and draft sidecars) and 85 aliases. Run hipfire list -r for the live list.

Family	Pull tags
Qwen 3.8 dense	qwen3.8:27b (MQ4V2 default), qwen3.8:27b-mq4-xts (flagship), -mq4-xt, -mq4-pro; MQ3, MQ5 and MQ6 -xt / base / -pro tiers; drafts qwen3.8:27b-draft-mq3 … -mq6
Qwen 3.8 Flash	qwen3.8:flash-next (alias qwen3.8:flash; MQ4 XTS split base), -mq3-xts, -mq2-xts, -mq2-xts-4500
Qwen 3.6 35B-A3B	qwen3.6:35b-a3b (MQ4P default), -mq2, -mq3p, -mq4p, -mfp4, -mq4r, -mq4rs (symmetric experts, gfx1201 fast prefill), -mq5, -mq6
Qwen 3.6 dense	qwen3.6:27b, qwen3.6:27b-mq3, drafts qwen3.6:27b-draft, -draft-mq3
Qwen 3.5 dense	qwen3.5:0.8b, 2b, 4b, 9b, 27b, with -mq3 / -mq6 variants and 9b / 27b drafts
Qwen 3.5 MoE	qwen3.5:35b-a3b
Qwen3	qwen3:0.6b, qwen3:8b
Muse Glimmer	muse-glimmer, muse-glimmer:fast (MQ4R), muse-glimmer:draft
Ornith 1.5	ornith-1.5:35b-a3b, ornith-1.5:fast (MQ4R)
DeepSeek V4 Flash	deepseek-v4-flash
MiniMax-M2.7	minimax-m2.7
North-Mini-Code-1.0	north-mini-code
Carnice (Hermes tool calls)	carnice:9b, carnice:27b, -mq6 variants
Qwopus	qwopus:4b, 9b, 27b, -mq6 variants, qwopus3.6:27b-coder
LFM2.5	lfm2.5:350m, 1.2b, 1.2b-thinking, 8b-a1b
NEX N2 Mini	nex-n2:mini
VibeThinker	vibethinker:3b, vibethinker:3b-mq6
FLUX (image)	flux.schnell:1 and others in IMAGEGEN.md

Curated weights are published under huggingface.co/hipfire-models. docs/MODELS.md covers sizes, minimum VRAM, sampling defaults, sidecars and provenance. To bring your own model, use hipfire quantize.

Qwen 3.8 Flash Next split-PLE tiers (0.4.3)

The hipfire-models/qwen3.8-flash-next repository publishes three plain split-PLE trunk files at its root, next to the shared PLE sidecar. The sizes below are file sizes in bytes, not VRAM or peak-memory requirements. Publishing a file is not a claim that every GPU route or speed has been measured on it. The published model card records the receipt-backed measurements and their acceptance scope.

Tier	File (HF link)	Bytes	SHA-256
MQ3	qwen3.8-flash-next-base.mq3-xts	59109362200	d35896ca69b87d3e93176847a036db3e2cb1554931f130a842b13718b486238b
MQ2 (6,306 hot)	qwen3.8-flash-next-base.mq2-xts	46895167000	e2ba747485190da92cd7a36d3f56809e6db31a3ec328f5ab23e000275d798e57
MQ2 (4,500 hot)	qwen3.8-flash-next-base.mq2-xts-4500	44537253400	9087d7d8f67629a3256c26bfeb705e7ed22dd1622bb91e074035543d6caea5c2
Shared PLE sidecar (Q4)	qwen3.8-flash-next.ple-q4	29120146432	fc78d30b9d2376db2b713732e80de2069f1cd1d6e1e40b50c716927dcc18ee1e

Trunk plus the Q4 sidecar totals 88,229,508,632 (MQ3), 76,015,313,432 (MQ2 6,306) and 73,657,399,832 (MQ2 4,500) bytes on disk. The MQ2 tiers need hipfire at or after ea2abd111f06339e9f25ac86715b7d210d0c5ae8; the MQ3 tier needs ca0aee9c925b4180651b194b440bcda80f656ac5.

MQ2 KLD on an RX 7900 XTX (gfx1100), post-land. Mean KLD against the FP8 reference, hipfire ea2abd111f06339e9f25ac86715b7d210d0c5ae8, with the shipping MQ2 hashes above. WikiText-2 (wt2): 32 chunks, 8160 scored rows; code24: 24 chunks, 6120 scored rows. Pre-land references describe earlier d2 predecessor bytes, not fresh measurements of the published roots; decode rows are paired against the weight-only reference and prefill rows against the prefill reference. The pre-land decode references (MQ2 6,306: 0.1851 wt2 / 0.1981 code24; MQ2 4,500: 0.2078 / 0.2077) are weight-only model KLD, not GPU route measurements; the XTX columns are GPU route measurements.

Tier	Corpus	Route	Pre-land reference	XTX post-land	Paired Δ (95% CI)	Relative Δ	Status
MQ2 6,306	wt2	decode	0.1851	0.1966	+0.0115 (+0.0075, +0.0155)	+6.23%	accepted by Kaden exact-row exception
MQ2 6,306	code24	decode	0.1981	0.2083	+0.0103 (+0.0018, +0.0160)	+5.18%	accepted, relative budget
MQ2 6,306	wt2	prefill	0.2095	0.2106	+0.0011 (−0.0058, +0.0080)	+0.55%	accepted, not significant
MQ2 6,306	code24	prefill	0.2111	0.2112	+0.0001 (−0.0059, +0.0068)	+0.05%	accepted, not significant
MQ2 4,500	wt2	decode	0.2078	not re-measured (identity + G2 pass)	—	—	not re-measured (identity + G2 pass)
MQ2 4,500	code24	decode	0.2077	0.2192	+0.0116 (+0.0053, +0.0179)	+5.56%	accepted, relative budget
MQ2 4,500	wt2	prefill	0.2274	0.2291	+0.0017 (−0.0055, +0.0091)	+0.76%	accepted, not significant
MQ2 4,500	code24	prefill	0.2237	not re-measured (identity + G2 pass)	—	—	not re-measured (identity + G2 pass)

Six of the eight cells were measured on the XTX. The MQ2 6,306 wt2 decode row (0.1966, +6.23%) exceeds the 6% relative budget and is accepted only as an explicit exact-row exception by Kaden; it is not a budget pass. The two remaining MQ2 4,500 cells were not re-measured. Product identity compared the builds on earlier d2 bytes (32 identical greedy token ids); G2 AR-lane parity passed on the published root hashes. Neither is a KLD measurement. This is not an all-eight-measured result. MQ3 XTX evidence is refusal-only; no XTX KLD or speed measurement is claimed.

GPU support

Family	Architectures	Notes
Vega / CDNA	gfx906, gfx908, gfx940–gfx942	Wave64 HIP; gfx906 dp4a MMQ prefill; some formats use a per-token GEMV fallback
RDNA1	gfx1010–gfx1013	Portable HIP and Redline dispatch
RDNA2	gfx1030–gfx1032	Portable HIP and Redline dispatch
RDNA3	gfx1100–gfx1103	WMMA kernels, validated MQ4R Redline route
RDNA3.5	gfx1150–gfx1152	WMMA kernels; Strix Halo (gfx1151) XDNA2 NPU prefill route
RDNA4	gfx1200, gfx1201	WMMA kernels, retained-PM4 Redline route, EP2 multi-GPU

Kernels are chosen through typed per-architecture dispatch tables. A specialization that is not available falls back to the portable or architecture fallback kernel. It is never borrowed from a neighbouring GPU family. Support for a GPU family does not mean every optimized route is admitted on it.

Redline

Redline is hipfire's in-tree dispatch and retained-replay layer. It records the actual kernel graph, derives resource dependencies, retains invariant command state, and lowers validated paths through public ROCr queue interfaces. If a graph is unsupported, shadow validation fails, the ABI does not match, a queue faults or the model changes, Redline falls back to ordinary HIP dispatch.

The Golden Redline guide reproduces the sealed MQ4R fixture (python3 -m tools.redline golden). Historical MQ4R measurements are in the gfx1201 campaign report. For the integration boundary, see HIPFIRE-GRAFT.md.

PeaceMaker

PeaceMaker is hipfire's own kernel toolchain. It builds AMD GPU and XDNA2 NPU programs directly, without hipcc. It works on typed machine code, and it can lift existing code objects so they can be checked against what PeaceMaker emits. It has six crates:

Crate	Role
peacemaker-ir	Typed RDNA machine-code representation and instruction tables
peacemaker-lift	Lifter for ELF, offload bundles, descriptors and metadata. Decoded kernels are re-encoded from typed fields and compared word for word
peacemaker-isa	Kernel builder and emitter (peacemaker-isa binary), with region import, wait-ledger checks and certification
peacemaker-author	PM kernel language (milestone M0, experimental, build-time only)
peacemaker-rip	Standalone .rip front end that lowers Rust-shaped kernel scripts onto the builder, without rustc
peacemaker-npu	XDNA2 / AIE2P program builder: encoder/scheduler, CDO/PDI and DPU transaction writers, CPU simulator

In 0.4.2, more production routes run builder-emitted, M7-certified PeaceMaker kernels. These ship as embedded kernels/*.hxaco modules with shape contracts. The hipcc versions of those kernels are kept only as oracle twins for exactness tests, not as runtime fallbacks. Examples from the CHANGELOG:

• sampled DFlash device acceptance on gfx1201
• CPU-AR sampled target rows and ordered acceptance on gfx1151 and gfx1201
• Qwen3.8-Flash EP owner assignment on gfx1201
• the gfx1151 MTP prompt-fill projections

The Strix Halo NPU prefill route builds its programs with peacemaker-npu. The offline railgun-cert recording-inventory gate keeps the list of recording and capture predicates in sync with the source tree. A kernel route with a certified PeaceMaker kernel is still subject to hipfire's normal admission and validation gates.

PeaceMaker is also published as a standalone repository, github.com/warpfront/peacemaker. scripts/mirror-push.sh peacemaker builds a history-preserving split of the PeaceMaker crates and their path dependencies. It then checks that the split builds and tests on its own, checks author identities and runs the publication guard on every commit. Only then does it push to main. Like hipfire, PeaceMaker is Apache-2.0. Some files remain MIT under their per-file SPDX header, and the mirror's LICENSE adds the LLVM exception notice (see scripts/mirror_split.py).

Install

Linux with ROCm 6 or newer:

curl -fsSL https://raw.githubusercontent.com/warpfront/hipfire/v0.4.3/scripts/install.sh | bash

# Or install the integration branch for testing:
curl -fsSL https://raw.githubusercontent.com/warpfront/hipfire/v0.4.3/scripts/install.sh \
  | bash -s -- --branch beta

RDNA4 requires ROCm 6.4 or newer, and gfx1151 requires ROCm 7.2 or newer. Release-tag installs download a verified prebuilt kernel pack for the detected arch when one is available. Otherwise they build kernels locally with hipcc. See Prebuilt kernel packs.

Before the first pull, hipfire doctor checks the driver and ROCm install, and hipfire fit <model> --ctx N says whether that model loads at that context on this machine, what the load would choose, and what to change if it does not.

hipfire version compares the installed binary, the managed source checkout and the daemon. To switch revisions, use hipfire update @beta, hipfire update --tag <tag> or hipfire update --commit <sha>. To uninstall while keeping models and settings:

curl -fsSL https://raw.githubusercontent.com/warpfront/hipfire/v0.4.3/scripts/uninstall.sh | bash

--dry-run previews the uninstall. --purge also deletes ~/.hipfire and asks for confirmation first.

For Windows, source builds and install verification, see docs/GETTING_STARTED.md. NixOS has a flake and a module (docs/NIXOS.md):

nix develop github:warpfront/hipfire  # dev shell with Rust + ROCm
nix build github:warpfront/hipfire    # build package

Containers (docs/CONTAINER.md):

docker compose up -d      # or: podman compose up -d
curl http://127.0.0.1:11435/v1/models

The image is ghcr.io/warpfront/hipfire:latest (docker pull ghcr.io/warpfront/hipfire:0.4.3.1); HIPFIRE_IMAGE=ghcr.io/warpfront/hipfire:0.4.3.1 pins compose to a release.

Deprecated features

PFlash prefill compression, CASK / TriAttention KV eviction, Givens/asymmetric KV modes and the legacy DFlash knobs have been deprecated since 0.4.0. They are unsupported and will be removed in 0.5.0. See configuration for supported settings and validation for the current gates.

Documentation

Page	Topic
GETTING_STARTED.md	Install, first run, what to read next
PICK-A-MODEL.md	Model recommendations by GPU tier
CLI.md	Every subcommand, flags, file locations
MODELS.md	Curated tags, BYO models, file extensions
QUANTIZE.md	hipfire quantize for HF / safetensors
CONFIG.md	Every config key and env override
SERVE.md	OpenAI-compatible HTTP API
multi-gpu.md	EP, TP and pipeline parallelism; budgets and refusals
IMAGEGEN.md	FLUX image generation
ARCHITECTURE.md	Engine layout and dispatch
QUANTIZATION.md	MQ formats, KV cache modes, FWHT
BENCHMARKS.md	Admitted measurements and claim requirements
VALIDATION.md	Correctness and speed gates
methodology/perf-benchmarking.md	Benchmark protocol
peacemaker-isa map / peacemaker-author map	PeaceMaker builder and kernel language
npu/README.md / peacemaker-npu map	XDNA2 NPU builder, simulator and tools
NIXOS.md / CONTAINER.md	NixOS and container builds

Contributing

See CONTRIBUTING.md, including the release cadence and PR cutoff. Install local hooks with ./scripts/install-hooks.sh. ./scripts/no-gpu-ci.sh runs the no-GPU CI subset, but it does not replace the hardware gates. No single gate establishes correctness on its own. Choose the route your change needs from docs/VALIDATION.md:

• scripts/serve_harness.py for generation and state-lifecycle changes.
• scripts/redline_daemon_harness.py for kernel, dispatch, graph or replay changes.

Do not bypass hooks with --no-verify.

Discord: https://discord.gg/F3BaywB8Rs

Acknowledgements

hipfire's gfx906 MMQ prefill path follows the design of the MMQ kernels in upstream llama.cpp. The community-maintained gfx906 forks of llama.cpp also informed that work. CREDITS.md lists contributors.

License

hipfire is licensed under Apache-2.0 as of v0.3.0. See LICENSE, LICENSE-APACHE and NOTICE. Releases up to and including v0.2.1 were dual MIT/Apache-2.0, and that grant is not revoked. Files whose authors have not elected Apache-2.0 remain MIT (LICENSE-MIT) and say so in their SPDX-License-Identifier. The decision record is docs/governance/relicense-2026-05.md. Architectural work that originated in hipfire is catalogued in PRIOR-ART.md. Derivative works should attribute it as described in AGENTS.md.