rootonchair/Lens-Turbo-GGUF

Lens-Turbo GGUF

GGUF quantizations of the Lens-Turbo diffusion transformer from microsoft/Lens-Turbo, prepared for use with stable-diffusion.cpp.

Lens-Turbo is the distilled 4-step variant of Microsoft Lens. These files contain only the diffusion model weights. For inference you also need:

• a GPT-OSS-20B text encoder GGUF, for example unsloth/gpt-oss-20b-GGUF
• the FLUX.2 VAE, for example from black-forest-labs/FLUX.2-dev

Files

File	Type	Approx. size	Notes
lens_turbo-f16.gguf	f16	7.7 GB	Highest precision GGUF export
lens_turbo-q8_0.gguf	q8_0	4.1 GB	High quality quantization
lens_turbo-q6_K.gguf	q6_K	3.2 GB	Strong quality/size balance
lens_turbo-q5_K.gguf	q5_K	2.7 GB	Smaller balanced option
lens_turbo-q4_K.gguf	q4_K	2.2 GB	Smallest provided option

Usage

Example with stable-diffusion.cpp:

./build/bin/sd-cli \
  --diffusion-model lens_turbo-q4_K.gguf \
  --llm /path/to/gpt-oss-20b-UD-Q8_K_XL.gguf \
  --vae /path/to/flux2_ae.safetensors \
  --vae-format flux2 \
  --cfg-scale 1.0 \
  --steps 4 \
  --diffusion-fa \
  -W 1024 \
  -H 1024 \
  -p "A crystal dragon soaring through an aurora borealis sky, its entire body made of transparent faceted crystal refracting the green and purple aurora light into rainbow spectra, ice particles trailing from its wings, high fantasy digital art"

If you keep the same folder structure as stable-diffusion.cpp examples, a local command can look like:

./build/bin/sd-cli \
  --diffusion-model models/lens_turbo-q4_K.gguf \
  --llm models/gpt-oss-20b-GGUF/gpt-oss-20b-UD-Q8_K_XL.gguf \
  --vae models/vae/diffusion_pytorch_model.safetensors \
  --vae-format flux2 \
  --cfg-scale 1.0 \
  --steps 4 \
  --diffusion-fa \
  -W 1024 \
  -H 1024 \
  -p "A crystal dragon soaring through an aurora borealis sky"

Download

Download one quantization:

hf download rootonchair/Lens-Turbo-GGUF lens_turbo-q4_K.gguf --local-dir models

Download all GGUF files:

hf download rootonchair/Lens-Turbo-GGUF --include "*.gguf" --local-dir models/Lens-Turbo-GGUF

Quantization Notes

The GGUF files were produced from the Lens-Turbo diffusion transformer weights using stable-diffusion.cpp convert mode:

./build/bin/sd-cli \
  -M convert \
  -m models/lens_turbo_bf16.safetensors \
  -o models/lens_turbo-q4_K.gguf \
  --type q4_K

The original sharded safetensors transformer was first merged into a single safetensors file before conversion.

No Q2 quantization is provided. In practice, Q2 is an extreme low-memory option and is more likely to noticeably degrade image quality. For most use cases, start with q4_K, q5_K, or q6_K; use q8_0 or f16 when quality is more important than size.

License and Attribution

This repository follows the upstream microsoft/Lens-Turbo license. Please review the original model card and license before using these weights.

Original model:

microsoft/Lens-Turbo

GGUF conversion:

rootonchair/Lens-Turbo-GGUF