image-gen-gguf — weight mirror

Files loaded by the local image generation feature of the KakaoWork desktop app, re-uploaded byte-identical from their upstream repositories. The app pins a sha256 for every file, so it needs a location whose bytes never change; this repository is that location. Nothing here is our own work.

Every upstream repository below is Apache-2.0 and ungated. Please prefer the upstream repositories — they hold the other quantizations, the model cards, and the people who actually did the work.

Files

File	Size	Upstream	sha256
qwen3-4b-instruct-2507-q4_k_m.gguf	2.50 GB	unsloth/Qwen3-4B-Instruct-2507-GGUF	3605803b982cb64aead44f6c1b2ae36e3acdb41d8e46c8a94c6533bc4c67e597
flux-ae-f16.gguf	0.17 GB	wbruna/Z-Image-Turbo-sdcpp-GGUF	1bed7b05318709e46a8cb9accc211168fc7f0b61ab594661860bbfe4d785cc46
z-image-turbo-q4_k_m.gguf	4.98 GB	jayn7/Z-Image-Turbo-GGUF	745ec270db042409fde084d6b5cfccabf214a7fe5a494edf994a391125656afd
flux2-klein-4b-q4_0.gguf	2.46 GB	leejet/FLUX.2-klein-4B-GGUF	d1023499ef3f2f82ff7c50e6778495195c1b6cc34835741778868428111f9ff4
qwen3-4b-q4_k_m.gguf	2.50 GB	unsloth/Qwen3-4B-GGUF	f6f851777709861056efcdad3af01da38b31223a3ba26e61a4f8bf3a2195813a
flux2-small-decoder.safetensors	0.25 GB	black-forest-labs/FLUX.2-small-decoder	ea4273f02d1fafbf8e1d1c2cf6018ed8748652eb0bf34f2dd91171f16f15ab62
z-image-turbo-q4_0.gguf	3.68 GB	leejet/Z-Image-Turbo-GGUF	2bc57986874c84f7ec6d02d9d7070a53b0029954a0e38a6e1342eb91095572f5
The list tracks exactly what the app loads. Weights for a preset we dropped
(Qwen Image Edit 13B) were removed in
867c70b.

Credits

Every file above is the work of its upstream author; we only copied the bytes. The underlying models are by their respective creators — Alibaba/Qwen (Qwen3), Black Forest Labs (FLUX.2), and Tongyi (Z-Image) — and the GGUF conversions are by the repositories listed in the table.

If you maintain one of the upstream repositories and would rather this mirror not exist, open a discussion and we will remove it.