Qwen3-4B (q4f16_1, WebLLM)

models.skillsafe.ai/qwen3-4b-q4f16@a5c9fab855e3/

Vetted New WebGPU

Text generation model for web-llm (q4f16_1). Runs on WebGPU — 2164 MB downloaded once from models.skillsafe.ai, then cached for every SkillSafe app that uses it. Inference happens on your device; nothing you enter is uploaded to load it, and it never costs a credit. Weights from Hugging Face · skillsafe-ai/qwen3-4b-q4f16, pinned at e37bb3fe8372 — also loadable straight from Hugging Face outside SkillSafe (how).

text generation web-llm WebGPU q4f16_1 Apache-2.0
Use it in an app Browse all models Request a variant
Share

Details

Catalogue idqwen3-4b-q4f16@a5c9fab855e3
Tasktext generation
Runtimeweb-llm ≥ 0.2.80
DeviceWebGPU
Variantq4f16_1
Download2164 MB · 75 files
LicenceApache-2.0 · notice
SourceHugging Face · skillsafe-ai/qwen3-4b-q4f16
Pinned ate37bb3fe8372610ddaaa6b824ff881266ac54b26
ApprovedSep 22, 2026
Statusactive — every file is a live vetted hash
API/v1/browser-models/qwen3-4b-q4f16

Files

Each file is served at an immutable URL; the canonical form is the SHA-256 itself. Tokenizer and config JSON are not here by design — they ship in your app bundle.

PathFormatSizeSHA-256
params_shard_0.bin bin 185 MB 961be4c2aa04…fce206
params_shard_1.bin bin 23.2 MB 687cd132c904…eb9501
params_shard_10.bin bin 23.8 MB 1dfaab1f0c12…28977d
params_shard_11.bin bin 30.4 MB ef41053e53d9…bf1af4
params_shard_12.bin bin 23.8 MB efadc3808019…92a87f
params_shard_13.bin bin 30.4 MB e8f4a78b6b02…b449c4
params_shard_14.bin bin 23.8 MB 03b2d91cf009…25b57d
params_shard_15.bin bin 23.8 MB 44de305eea80…ff99be
params_shard_16.bin bin 30.4 MB 70e1448b554c…91c5ed
params_shard_17.bin bin 23.8 MB 5a8f14fa5037…a751f7
params_shard_18.bin bin 30.4 MB 33480b3e8fb5…179335
params_shard_19.bin bin 23.8 MB 34bc7fe5f9a0…9231e4
params_shard_2.bin bin 23.8 MB 24c47c8e3290…e3b79a
params_shard_20.bin bin 30.4 MB 3ecd75b8123f…2d96d4
params_shard_21.bin bin 23.8 MB c8e0cba74578…418d69
params_shard_22.bin bin 30.4 MB ef483bc20554…338318
params_shard_23.bin bin 23.8 MB 92cadd88b95e…27c655
params_shard_24.bin bin 30.4 MB 76ed233293e3…c31396
params_shard_25.bin bin 23.8 MB 269df217a678…06ffd6
params_shard_26.bin bin 30.4 MB fd20afd6241b…574a5f
params_shard_27.bin bin 23.8 MB d8f009f1ade6…1deba3
params_shard_28.bin bin 30.4 MB dd226d8c5da6…0859e0
params_shard_29.bin bin 23.8 MB 68a1b731f824…8b461c
params_shard_3.bin bin 30.4 MB f11f98764913…33bd16
params_shard_30.bin bin 30.4 MB 7da28c17667c…e51f99
params_shard_31.bin bin 23.8 MB ef0880eb6a2f…6955df
params_shard_32.bin bin 30.4 MB aa5a0953f0f7…64d77c
params_shard_33.bin bin 30.4 MB 909155109d7d…4088cb
params_shard_34.bin bin 23.8 MB c9a99fe77eee…11b231
params_shard_35.bin bin 30.4 MB 1cc78e900ded…2e8f10
params_shard_36.bin bin 23.8 MB aa2d93416d95…7f8360
params_shard_37.bin bin 30.4 MB 97a0ab806a6f…506584
params_shard_38.bin bin 23.8 MB 7554d4c7c733…2cd1cf
params_shard_39.bin bin 30.4 MB de669be3f5ed…31e33b
params_shard_4.bin bin 23.8 MB dd2d3c137376…0c0e84
params_shard_40.bin bin 23.8 MB 64d6751289f6…34842b
params_shard_41.bin bin 30.4 MB f27d8720407c…804604
params_shard_42.bin bin 23.8 MB 87158d3140be…157bae
params_shard_43.bin bin 30.4 MB 87035a905e57…c6b5d3
params_shard_44.bin bin 23.8 MB 0a440e2eca28…ad2e8a
params_shard_45.bin bin 30.4 MB 65f0d32d175d…9aa646
params_shard_46.bin bin 23.8 MB 7b2a977e84b8…2a831b
params_shard_47.bin bin 30.4 MB 7c08681f6bf7…201513
params_shard_48.bin bin 23.8 MB 0fc5bb0e8d81…9f2f03
params_shard_49.bin bin 30.4 MB 6171c7b76f39…3a122e
params_shard_5.bin bin 30.4 MB 5fee166de4ad…09f71c
params_shard_50.bin bin 23.8 MB c8c08adacc71…af9247
params_shard_51.bin bin 30.4 MB 306900717253…e84081
params_shard_52.bin bin 23.8 MB 51bb4dc0af4b…8d860e
params_shard_53.bin bin 30.4 MB f8ef3afe217c…8f9957
params_shard_54.bin bin 23.8 MB 14c76365a17a…a0922f
params_shard_55.bin bin 30.4 MB 3a5e4ef769df…c70401
params_shard_56.bin bin 23.8 MB 8b9f098eab92…5b5c2f
params_shard_57.bin bin 30.4 MB 0bf24f338ed4…422566
params_shard_58.bin bin 23.8 MB 7fae8ca8360a…dad47c
params_shard_59.bin bin 30.4 MB 7e84743390ba…49eb33
params_shard_6.bin bin 23.8 MB eb5ed96c8d57…cdc3b6
params_shard_60.bin bin 23.8 MB 40044b185e03…7c6050
params_shard_61.bin bin 30.4 MB 30a8d4d355ce…5e0615
params_shard_62.bin bin 23.8 MB 5526d1f267e4…fd34ca
params_shard_63.bin bin 30.4 MB 8f207d6d10ef…369beb
params_shard_64.bin bin 23.8 MB ef345f8cd4a5…545ae6
params_shard_65.bin bin 30.4 MB f1d2228b3f05…6b4483
params_shard_66.bin bin 23.8 MB 7564658ad7ad…6024c2
params_shard_67.bin bin 30.4 MB 00f54fac326a…e9f94d
params_shard_68.bin bin 23.8 MB 9e0908e2ea2e…77008d
params_shard_69.bin bin 30.4 MB ffbddb87d1a9…04891c
params_shard_7.bin bin 30.4 MB cb8319bc8cdd…ea56f0
params_shard_70.bin bin 23.8 MB ec1800cee1e7…b134ee
params_shard_71.bin bin 23.8 MB 4a75e710f170…d7a1de
params_shard_72.bin bin 30.4 MB 4cc8468ae134…f629aa
params_shard_73.bin bin 30.4 MB b82c7a5f80b2…c16a94
params_shard_8.bin bin 23.8 MB 73b841db0959…25b276
params_shard_9.bin bin 30.4 MB f76e18c570aa…093bea
lib/Qwen3-4B-q4f16_1-ctx4k_cs1k-webgpu.wasm wasm 5.8 MB 5ddf44e49b03…571340
Use it in an app declaration · SDK loader · web-llm · URLs · Hugging Face — generated from this entry

Add to the body of POST /v1/apps/{slug}/releases (or a release session). An unknown or withdrawn model is refused with a 400 naming it; the app page then shows "downloads 2164 MB · runs on your device" and /models.txt carries the attribution.

{
  "models": [
    {
      "id": "qwen3-4b-q4f16",
      "revision": "a5c9fab855e3"
    }
  ]
}

Evaluation

shard sizes + md5 vs ndarray-cache.json; implied parameter count vs the checkpoint; parameter table vs model_lib embedded metadata — MLC's published weights, checked against the WebGPU library. Evaluated Sep 22, 2026.

Library fit

params_matched: 435 · symbolic_dims_resolved: [] · model_type: qwen3 · quantization: q4f16_1 · effective: {"context_window_size":4096,"prefill_chunk_size":1024}

Toolchain: python 3.12.13 · platform Darwin 25.6.0 arm64 · torch 2.10.0 · onnx 1.23.0 · onnxruntime 1.30.0. Recipe models/recipes/qwen3-4b-q4f16.yaml (1639cc66dc94). Full manifest.json

Licence & attribution

Apache-2.0 · licence text · notice

Qwen3-4B: Copyright 2025 Alibaba Cloud. Apache License 2.0. https://huggingface.co/Qwen/Qwen3-4B — q4f16_1 weights published by MLC (https://huggingface.co/mlc-ai/Qwen3-4B-q4f16_1-MLC).

Apps that declare this model get this text in their generated /models.txt, so a licence that requires a notice always carries one.

Revisions

RevisionVariantSizeApprovedStatusApps
qwen3-4b-q4f16@a5c9fab855e3 q4f16_1 2164 MB Sep 22, 2026 active 0
qwen3-4b-q4f16@a5c9fab8 q4f16_1 2158 MB Sep 22, 2026 withdrawn 0

Every file here was approved by exact SHA-256 after a structural audit of the graph, fetched from a content-pinned source, and is served credential-free at an immutable URL. The SDK re-verifies the hash on your device before it caches or returns anything. Missing a variant? Request it — or read how the registry works.