Quantized, browser-streamable repacks of FLUX.2 [klein] 4B (rectified-flow image transformer + VAE) and Qwen3-4B (as the text encoder — hidden-state taps at layers 9/18/27), packed for fully client-side WebGPU inference: no server-side compute, weights fetched with plain HTTP GETs and streamed straight into GPU buffers.
| bundle | contents | size | quantization |
|---|---|---|---|
flux2-q8/ |
FLUX.2 klein transformer + VAE | ~4.0 GB | int8 g64 matmuls, f16 rest |
qwen3-q8/ |
Qwen3-4B layers 0–26 + tokenizer | ~3.2 GB | int8 g64 |
qwen3-q4/ |
Qwen3-4B layers 0–26 + tokenizer | ~1.85 GB | int4 g64, int8 embeddings |
gdino-q8/ |
Grounding DINO base (full detector) | ~328 MB | int8 g64, Swin at f16 |
sam2-q8/ |
SAM 2.1 hiera-small (image path) | ~41 MB | int8 g64 |
The bundle layout is purpose-built for streaming into WebGPU (not a
standard interchange format like safetensors or GGUF): each bundle
directory holds a manifest.json (tensor table, shard offsets,
tokenizer/chat-template segments) plus shard_NNN.bin files (~250 MB
each) — CDN/CORS friendly, cacheable as immutable content.
Validated end-to-end against the original-weights reference implementation (fp32 eager) on identical inputs:
An int4 variant of the image transformer failed these gates (velocity cosine 0.96) and is deliberately not published.
A consuming runtime needs WebGPU with ~8 GB of GPU/unified memory for
flux2-q8 + qwen3-q4; shader-f16 is used when available. Validated on
Chrome (Vulkan, D3D12, and Metal backends) and Safari 26+.
Character-reference conditioning (multi-image) is supported natively
by the packed transformer weights.
Both source models are licensed under the Apache License 2.0; these repacks are derivative works distributed under the same license. See NOTICE for required attributions and the full statement of modifications (quantization parameters, omitted layers, format changes).
Not affiliated with or endorsed by Black Forest Labs Inc., Alibaba Cloud, IDEA-Research, or Meta Platforms. "FLUX" is a trademark of Black Forest Labs Inc., used here only to describe the origin of the weights.
qwen3-q4/shard_000.bin 401 MBqwen3-q8/shard_000.bin 401 MBqwen3-q4/shard_005.bin 268 MBqwen3-q4/shard_001.bin 268 MBqwen3-q4/shard_002.bin 268 MBqwen3-q4/shard_004.bin 268 MBqwen3-q4/shard_003.bin 268 MBgdino-q8/shard_000.bin 267 MBflux2-q8/shard_015.bin 264 MBflux2-q8/shard_001.bin 263 MBflux2-q8/shard_005.bin 263 MBqwen3-q8/shard_007.bin 262 MBqwen3-q8/shard_002.bin 261 MBqwen3-q8/shard_004.bin 261 MBqwen3-q8/shard_010.bin 261 MBqwen3-q8/shard_006.bin 261 MBflux2-q8/shard_000.bin 260 MBqwen3-q8/shard_009.bin 260 MBqwen3-q8/shard_001.bin 260 MBqwen3-q8/shard_003.bin 260 MBqwen3-q8/shard_005.bin 260 MBqwen3-q8/shard_008.bin 258 MBflux2-q8/shard_002.bin 253 MBflux2-q8/shard_003.bin 253 MBflux2-q8/shard_004.bin 253 MBflux2-q8/shard_006.bin 253 MBflux2-q8/shard_007.bin 253 MBflux2-q8/shard_008.bin 253 MBflux2-q8/shard_009.bin 253 MBflux2-q8/shard_010.bin 253 MBflux2-q8/shard_011.bin 253 MBflux2-q8/shard_012.bin 253 MBflux2-q8/shard_013.bin 253 MBflux2-q8/shard_014.bin 253 MBqwen3-q8/shard_011.bin 208 MBqwen3-q4/shard_006.bin 107 MBflux2-q8/shard_016.bin 106 MBgdino-q8/shard_001.bin 61 MBsam2-q8/shard_000.bin 41 MBqwen3-q8/check_hs0.bin 1 MBqwen3-q8/check_hs1.bin 1 MBqwen3-q8/check_hs18.bin 1 MBqwen3-q8/check_hs27.bin 1 MBqwen3-q8/check_hs9.bin 1 MBHave the files? Submit a torrent.
Submit a torrentHave the complete files and permission to share them? Replace the example folder path. The script checks the files, makes a torrent and submits its magnet and checksums. Keep your torrent client seeding. Get your community key from your settings, and read the script before running it.
curl -fsSL https://pirateface.co/package.sh | bash -s -- --repo gandamu/flux2-klein-4b-qwen3-4b-webgpu ./model-folderIDEA-Research/grounding-dino-base