Phi-3.5-mini-instruct (q4f16_1, WebLLM)

models.skillsafe.ai/phi-3.5-mini-instruct-q4f16@206a7da2/

Vetted New WebGPU

Text generation model for web-llm (q4f16_1). Runs on WebGPU — 2055 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/phi-3.5-mini-instruct-q4f16, pinned at 1bdc05b138f9 — also loadable straight from Hugging Face outside SkillSafe (how).

Share

Details

Catalogue idphi-3.5-mini-instruct-q4f16@206a7da2
Runtimeweb-llm ≥ 0.2.80
DeviceWebGPU
Variantq4f16_1
Download2055 MB · 84 files
LicenceMIT · notice
Pinned at1bdc05b138f907970e0c795916e9dd067c058623
ApprovedSep 22, 2026
Statusactive — every file is a live vetted hash

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 47.0 MB 8af692b159dd…8cfafe
params_shard_1.bin bin 24.0 MB e3058b06b585…49302a
params_shard_10.bin bin 20.3 MB 434f67218aca…eea537
params_shard_11.bin bin 24.0 MB 6f50bd6ea725…ac5b17
params_shard_12.bin bin 21.6 MB 752412dde2ab…967410
params_shard_13.bin bin 24.0 MB cb6bcd252c55…1364f0
params_shard_14.bin bin 31.7 MB 481d8c33b6e5…0eecf2
params_shard_15.bin bin 20.3 MB 4f874c64a10f…4c5f5c
params_shard_16.bin bin 24.0 MB 39763e6f7783…70a502
params_shard_17.bin bin 21.6 MB c01663dbd88d…8d501c
params_shard_18.bin bin 24.0 MB 7d4e6c823be8…6748b0
params_shard_19.bin bin 31.7 MB 3a89d5823972…068ba9
params_shard_2.bin bin 22.4 MB 48cca340f994…91b371
params_shard_20.bin bin 20.3 MB 5fc4ac5bea34…f9d7e4
params_shard_21.bin bin 24.0 MB 3270587fc5ca…03f302
params_shard_22.bin bin 21.6 MB 91cce7d0c10e…cfe46e
params_shard_23.bin bin 24.0 MB e9b35f01157a…a42458
params_shard_24.bin bin 31.7 MB 8b6316bc5c47…06e894
params_shard_25.bin bin 20.3 MB b2e4aeae8aff…15ab3f
params_shard_26.bin bin 24.0 MB 69206ca75165…a16b43
params_shard_27.bin bin 21.6 MB abcc7b5b9054…f021c8
params_shard_28.bin bin 47.0 MB 89a7a3d48031…740037
params_shard_29.bin bin 21.1 MB 220c8f229827…92d651
params_shard_3.bin bin 24.0 MB 3e93dd428ad0…38baef
params_shard_30.bin bin 24.0 MB 5857bf34ddc6…cbf0ef
params_shard_31.bin bin 21.6 MB 21e012d5b8a5…eb1bc6
params_shard_32.bin bin 24.0 MB fc8e92e2a4d3…f06dea
params_shard_33.bin bin 31.7 MB 5cdb872295a5…8d9c5a
params_shard_34.bin bin 20.3 MB 057aff7d53cd…32c4a5
params_shard_35.bin bin 24.0 MB b436736d30f5…5be4c7
params_shard_36.bin bin 21.6 MB dc2aaff67064…f19358
params_shard_37.bin bin 24.0 MB 2245d38a540e…a9f269
params_shard_38.bin bin 31.7 MB 788a86e9c2e7…c456e9
params_shard_39.bin bin 20.3 MB cafcc2839907…45366a
params_shard_4.bin bin 31.7 MB d6b88da8a732…dc759e
params_shard_40.bin bin 24.0 MB afa6f924090e…cfc457
params_shard_41.bin bin 21.6 MB d0c2ffa02577…44be94
params_shard_42.bin bin 24.0 MB 96eb08889569…4ea755
params_shard_43.bin bin 31.7 MB fa4473ade710…d25881
params_shard_44.bin bin 20.3 MB 42a66fe11b1b…1abaa6
params_shard_45.bin bin 24.0 MB a3aded3a24c8…d8d117
params_shard_46.bin bin 21.6 MB 9491fe711a65…820b10
params_shard_47.bin bin 24.0 MB 9aeee1cfee83…f1d3dc
params_shard_48.bin bin 31.7 MB 02b1c55e9caa…8f5cb4
params_shard_49.bin bin 20.3 MB addaad6e1aeb…4352df
params_shard_5.bin bin 20.3 MB 2aaffc75b728…3ecbef
params_shard_50.bin bin 24.0 MB 487ef5593c09…04fd3a
params_shard_51.bin bin 21.6 MB 87a1c16232bb…218173
params_shard_52.bin bin 24.0 MB d004ac0b4537…caa153
params_shard_53.bin bin 31.7 MB f6d5f005600c…08a187
params_shard_54.bin bin 20.3 MB d1c2024d2d53…f1dca5
params_shard_55.bin bin 24.0 MB b847525d1ef6…828f4d
params_shard_56.bin bin 21.6 MB 082000be586a…0feac9
params_shard_57.bin bin 24.0 MB 83ba70fb8d4e…a70d89
params_shard_58.bin bin 31.7 MB fb00013e1220…f5e53c
params_shard_59.bin bin 20.3 MB 761be683bb52…c456f4
params_shard_6.bin bin 24.0 MB 3dd99d664355…992944
params_shard_60.bin bin 24.0 MB 09eea2057bdb…c86657
params_shard_61.bin bin 21.6 MB 19cbd4aa4ff8…3ed62e
params_shard_62.bin bin 24.0 MB 008e98ce3331…d60f3e
params_shard_63.bin bin 31.7 MB c7e7aa058469…a78890
params_shard_64.bin bin 25.3 MB beaa2570c785…038fa0
params_shard_65.bin bin 24.0 MB 30c3c01277d0…1920b0
params_shard_66.bin bin 21.6 MB 6b07d8090e47…ca7e4b
params_shard_67.bin bin 24.0 MB 19d6885e5c43…dfbfaf
params_shard_68.bin bin 31.7 MB 42deffa89fd9…4dc46c
params_shard_69.bin bin 20.3 MB 44c670e51da3…832447
params_shard_7.bin bin 21.6 MB 0e4ddc40d122…16b81c
params_shard_70.bin bin 24.0 MB 63837c95ae6f…39b474
params_shard_71.bin bin 21.6 MB 7706a7477f12…2db2fe
params_shard_72.bin bin 24.0 MB 1742d7e60759…62cd76
params_shard_73.bin bin 31.7 MB 4064341be386…8c67df
params_shard_74.bin bin 20.3 MB 5e080b00cb28…679b2c
params_shard_75.bin bin 24.0 MB 25f9a28a7a59…4bfa54
params_shard_76.bin bin 21.6 MB 5524164b8caa…369539
params_shard_77.bin bin 24.0 MB 1057d60644b6…f64abe
params_shard_78.bin bin 31.7 MB 049bd9abd6e3…80d796
params_shard_79.bin bin 20.3 MB 044ae09eb48e…5326f2
params_shard_8.bin bin 24.0 MB 847b463dfa95…ed0c79
params_shard_80.bin bin 24.0 MB c546de35cc45…0cacc5
params_shard_81.bin bin 21.6 MB a57b09c0005d…69f7c5
params_shard_82.bin bin 15.2 MB 8a054211e642…9dd878
params_shard_9.bin bin 31.7 MB db2beb4d7b3f…47bd2a
lib/Phi-3.5-mini-instruct-q4f16_1-ctx4k_cs1k-webgpu.wasm wasm 5.2 MB 2f58ffc1ed1d…2baa6f
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 2055 MB · runs on your device" and /models.txt carries the attribution.

{
  "models": [
    {
      "id": "phi-3.5-mini-instruct-q4f16",
      "revision": "206a7da2"
    }
  ]
}

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: 325 · symbolic_dims_resolved: ["lm_head.q_scale","lm_head.q_weight"] · model_type: phi3 · 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/phi-3.5-mini-instruct-q4f16.yaml (5abf8d42827f). Full manifest.json

Licence & attribution

MIT · licence text · notice

Phi-3.5-mini-instruct: Copyright Microsoft, MIT License. https://huggingface.co/microsoft/Phi-3.5-mini-instruct — q4f16_1 weights published by MLC (https://huggingface.co/mlc-ai/Phi-3.5-mini-instruct-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.

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.