Models

Owner: registry-backed model surface (docs/INDEX.md). Machine sources: curated registry/models.json; generated and bundled registry/v1.json (loaded by hipfire-registry). Last checked: 2026-08-24 against the published Qwen3.8 MQ V2 ladder.

This page projects registry availability: tags, default artifact filenames, declared download size, and declared VRAM floor. It is not a product admission table and not a guarantee that every GPU/route runs every tag.

hipfire list -r prints the live registry plus local availability. Prefer that command when sizes change; this page is a checked narrative, not a second registry.

Pull and run

hipfire pull qwen3.5:9b
hipfire run qwen3.5:9b "hello"
hipfire list -r

Default serve pre-warm tag is qwen3.5:9b (CONFIG.md → default_model). Per-tag sampling defaults come from registry recommended_settings only (applied by the native CLI request resolver). Cards may set intentional reasoning_effort (semantic prompt strength only — never a budget). Effort-native families (Qwen3.8, Ornith 1.5, DeepSeek V4, Muse Glimmer) omit thinking_budget; absence means no implicit cap. Qwen3.8 and Ornith 1.5 accept hipfire’s explicit integer Qwen continuation cap. DeepSeek V4 and Muse Glimmer cap fields are dropped+warned rather than translated into hand-written tokenizer closure. Registry sampling blocks are legacy metadata and are not promoted by the native request resolver. Full contract: CONFIG.md; HTTP examples: SERVE.md.

Registry tags (from registry/models.json)

Fields: Tag, File (file), Size GB (size_gb), Min VRAM GB (min_vram_gb), Default KV (default_kv_mode when set; else empty — global kv_cache=auto resolves to q8), Notes (desc, truncated).

Qwen 3.5 dense / hybrid

Qwen 3.5 / 3.6 MoE (A3B)

Sizes below are registry declarations, not a substitute for runtime MoE layout checks.

Several A3B entries carry an mtp.file sidecar name (qwen3.6-35b-a3b.mtp). MTP enablement is config/runtime gated (mtp_mode, env); registry presence alone is not admission.

Qwen 3.6 dense

Qwen 3.8 dense

MQ2V2 is not registered. Explicit qwen3.8:27b-mq4 aliases to qwen3.8:27b. Legacy qwen3.8:27b-fast / qwen3.8:fast alias to qwen3.8:27b-mq4-xt. Reasoning contract: CONFIG.md (effort is semantic only; named thinking_budget dropped on this family).

DFlash draft artifacts (registry)

Draft loading is registry-driven: hipfire pull <tag> fetches the draft sidecar alongside the target, and dflash_mode / speculation (CONFIG.md, env-vars.md) decides what happens next — default dflash_mode is off (pull ≠ enable); auto uses the sidecar when present (AR otherwise); on requires it and fails the load without it. Override the sidecar with developer.dflash_draft / HIPFIRE_DFLASH_DRAFT or run --model-draft. Watch for DFlash draft loaded: in load output. Several tags may share one dflash.file; hipfire rm keeps that file while another installed declarer still needs it (CLI.md).

Vision-tower loading is registry-driven the same way: every qwen3.8:27b* tier declares the shared vision.file (qwen3.8-27b-vision.hfq, llama.cpp mmproj-style), so hipfire pull <tag> fetches the tower once and every text quant tier serves images without requantizing the trunk. The loader opens the sidecar as a separate pack and applies it with the trunk’s vision config. vision_mode (CONFIG.md, env-vars.md) decides what happens next — default off never loads the tower (even an explicit --vision is skipped, the daemon’s dflash_mode=off-style hard override, so text loads pay no tower VRAM); auto uses the registry/sibling sidecar when present and stays silently text-only when absent; on requires the declared sidecar and fails the load closed without it. A trunk with an embedded tower declares no sidecar and is unaffected by this key. Override per load with run --vision / serve --vision or HIPFIRE_VISION_SIDECAR (empty opts out); a <trunk-stem>-vision.hfq file beside the trunk is also discovered. hipfire rm keeps the shared file while another installed declarer still needs it (CLI.md).

Qwen3 (non-3.5) dense HF4

Fine-tunes on Qwen 3.5 / 3.6 families

Other families (registry)

LFM2.5 (registry)

Registry recommended_settings for LFM tags is low temperature (0.05–0.2) with repeat_penalty 1.05 — applied by the CLI resolver. Do not treat a registry sampling field as active defaults.

Gemma4 (registry — no published artifact yet)

No hipfire-quantized Gemma4 .hfq/.mq4 artifact has been published. The rows below are the intended tag layout for when the quantize + upload lane publishes; the current registry/models.json intentionally lists zero gemma4: tags so that no registry row can point at a 404. See the report at the bottom of this section for the exact repos/files to publish.

Architecture ground truth (crates/hipfire-arch-gemma4/src/config.rs, crates/hipfire-arch-gemma4/src/lowered.rs, and crates/hipfire-arch-gemma4/src/gemma4.rs):

• Hybrid 5:1 sliding:global — every 6th layer is global (full) attention. The per-layer layer_types array is authoritative; on the 12B dense text config (hidden_size=3840, num_hidden_layers=48, vocab_size=262144) that is 40 sliding + 8 full. Do not assume the period in code.
• Sliding-window layers: window = 1024 (sliding_window), head_dim = 256 (sliding_head_dim / head_dim), RoPE θ = 10_000 (sliding_rope_theta), RopeType::Default on full head_dim, attention scale = 1.0 (kernels bake 1/√d so decode pre-scales Q by √head_dim), Q/K per-head RMSNorm.
• Global (full) layers: head_dim = 512 (global_head_dim / full_head_dim), window = 0 (full causal), RoPE θ = 1_000_000 (full_rope_theta), RopeType::Proportional with partial_rotary_factor = 0.25 (only first 64 of 512 dims rotate; remainder NoPE), attention_k_eq_v = true — V shares the pre-k_norm output of k_proj (no v_proj on those layers; v_norm is weight-less, implemented with a ones-filled scratch buffer).
• KV head counts (variant-dependent): 12B text — num_attention_heads = 16, num_key_value_heads = 8 (sliding) and num_global_key_value_heads defaults to sliding when absent; 31B target (per lowered.rs comments) — n_heads = 32, sliding_n_kv_heads = 16, full_n_kv_heads = 4, sliding_head_dim = 256, full_head_dim = 512, hidden_dim = 21504. The 12B hidden_dim is intermediate_size = 15360, dim = 3840.
• FFN: SwiGLU with gelu_pytorch_tanh activation, intermediate_size (above) per-layer.
• Norms: Sandwich RMSNorm — input_layernorm, post_attention_layernorm, pre_feedforward_layernorm, post_feedforward_layernorm per layer plus a learned scalar layer_scalar [1] at layer end. Gemma4 uses plain x * w (weights init 1.0); the HIPFIRE_GEMMA4_NORM_PLUS_ONE=1 toggle bakes the Gemma-2/3 x * (1+w) form at load time.
• Embeddings: embed_scale = sqrt(hidden_size) multiplied on every lookup; lm_head is TIED to embed_tokens (single GPU allocation, aliased WeightTensor; free skips the head to avoid double-free).
• Output: final_logit_softcapping = 30.0 — tanh(logits/30)*30 before sampling. norm_eps = 1e-6, max_position_embeddings = 262144 (12B).
• Stop / EOS: HF eos_token_id is the list [1, 106] (<eos> = 1 and <end_of_turn> / <turn|> = 106); parsing it as a scalar drops 106 and lets decode loop on <turn|> forever. The loader resolves eos_tok against ["<end_of_turn>", …] and masks both.
• Sampling defaults (Gemma card): temperature 1.0, top_p 0.95, top_k 64 — the registry will carry these in recommended_settings / sampling_profiles when tags land (today no Gemma rows exist to carry them).
• Thinking: boolean only (official default off). No native reasoning_effort; unsupported effort/named-budget values are dropped with a warning. See CONFIG.md / SERVE.md.
• Crate: hipfire-arch-gemma4 (Gemma4Config, Gemma4Weights, Gemma4State), Gemma4Bundle { config, weights, state, eos_tok } in hipfire-loader. Gemma4Carrier claims arch ids 13 (gemma4_text) and 22 (gemma4_unified_assistant EAGLE drafter) — see architecture-ids.md.

No Gemma4 rows are registered until the files above (or the 27B/31B MoE variants) are actually on the Hub and pass scripts/registry_gen.py LFS probing — that script is the admission gate and will fail-closed on a missing file or mismatched size_gb. Publishing steps: hipfire quantize → upload to the Hub → add a models.json entry mirroring a dense neighbour like qwen3.6:27b (fields: repo, file, size_gb, min_vram_gb, desc, recommended_settings with temp 1.0/top_p 0.95/top_k 64) → run scripts/registry_gen.py to stamp sha256/size_bytes/arch_id/quant.

Aliases

String redirects in registry/models.json → aliases (not separate downloads). Partial table — for the complete surface read that file or run hipfire list -r.

Runtime family map (source, not registry)

Runtime dispatch uses HFQ arch_id (architecture-ids.md). Summary for operators:

Dense LFM2.5 is supported on arch_id 11. The LFM config parser treats num_experts == 0 as dense SwiGLU on every layer (crates/hipfire-arch-lfm2moe/src/config.rs). Do not claim dense LFM is unsupported.

hipfire-arch-lfm2moe is a non-optional dependency of hipfire-loader / daemon load paths on this tree (see crate Cargo.toml graphs). Feature flags on hipfire-runtime default set do not list a separate arch-lfm2moe toggle the way some other arches do — loader always links the crate.

Capability features (DFlash, CASK, PP, MTP, batched prefill, n-gram) are per-path and often narrower than “model loads”. Spec inventory history: speculation-support-inventory.md (historical). Product claims need source + admissions.yml.

LFM optimized prefill — branch-only scope

Branch-only; not shipped on origin/beta@202282de8759dfa6963ea5184ad2bf2b9259cef6.

Audited branch wording allowed for optimized LFM prefill (and nothing broader):

• Exact cohort: 350M dense MQ4 fixture path used by the branch runtime fixture validation/guard (lfm2.5-350m.mq4 shape checks in hipfire-arch-lfm2moe forward), not a generic “all LFM” claim.
• GPU: gfx1201 only for the batched opt-in path.
• Flag: explicit opt-in HIPFIRE_LFM2_PREFILL_BATCH=1 (default off). Optional chunk override HIPFIRE_LFM2_PREFILL_MAX_BATCH (default 256, hard cap 512 in source).
• Pin when citing branch implementation: lfm-redline@692a726dde53508cb53de1a74c720e75a7c9f33e (or later branch commits only if re-grounded).

Planned (not implemented claims here): Q8-first generic completion of the optimized path, wider LFM cohorts (1.2B / 8B-A1B), multi-GPU, and Phase-4 default-on. Admitted (exact one row): admissions.yml schema v2 admits only the sealed gfx1201 LFM2.5-350M MQ4 retained-PM4 plain-AR product route; nothing else. Not a current baseline: any exploratory tok/s tables in designs/plans.

Eager per-token prefill / decode remains the portable LFM path when the opt-in flag is off or the GPU is not gfx1201. On gfx1201 with HIPFIRE_LFM2_PREFILL_BATCH=1, the daemon selects the batched path from GPU+flag alone and has no post-selection fallback: requests outside the exact 350M dense MQ4 fixture fail closed at the runtime fixture guard. Source symbol validate_350m_mq4_admission names that fixture check only — it does not create a product admission; admissions.yml remains the sole authority (schema v2, exactly one earned retained-PM4 product row for this sealed fixture).

Bring your own

HuggingFace → quantize → register

hipfire quantize Jackrong/Qwopus3.5-4B-v3 \
  --format mq4 --install --register qwopus:4b

See QUANTIZE.md and QUANTIZATION.md.

Local safetensors directory

Requires config.json + .safetensors. Architectures the engine loads are those with arch crates / loaders above; the quantizer may accept more shapes than inference can run.

GGUF

hipfire quantize ./model.Q4_K_M.gguf --install --register my:tag

Dequant path support is format-specific (common Q4_0 / Q8_0 / Q4_K / Q6_K / F16 / BF16 / F32). Unsupported GGUF quants fail closed in the quantizer.

On-disk layout

~/.hipfire/models/
  <registry file names>
  optional sibling drafts / .triattn*.bin sidecars

Extension hints (loader recognizes several): .mq4, .mq6, .mq4p, .mq4r, .mq2, .mq2lloyd, .mq2r, .mfp4, .hf4, .hf6, .hfq, .q8, and related graded names as produced by quant tooling. Exact dtype routing is loader/kernel source, not this table.

Thinking / chat framing

Reasoning is a three-axis contract (mode / semantic effort / hard cap) owned by config and the OpenAI request layer — not by inventing per-tag field meanings:

• Axes, defaults, and family table — CONFIG.md
• HTTP fields, warn+drop metadata, curl examples — SERVE.md
• Chat template overrides — chat_template, default_chatml / env in env-vars.md

Registry cards may publish reasoning_effort defaults. Effort-native tags should not pin a hipfire named thinking_budget; absence means uncapped. Ornith 1.5 uses Qwen3.8-compatible low/medium/xhigh semantic prompt steering (default xhigh) — never a budget reinterpretation. Qwen <think> framing is not universal — Gemma uses boolean thinking (default off), Glimmer uses Onyx channel strength, DeepSeek uses thinking.type + its own effort ladder.

Related