'FLUX.2 Small Decoder' has been released, which speeds up the decoding process of the image generation AI 'FLUX.2' by 1.4 times.
AI development company Black Forest Labs released ' FLUX.2 Small Decoder, ' a high-speed decoder for the FLUX.2 series, on April 8, 2026.
black-forest-labs/FLUX.2-small-decoder · Hugging Face
Releasing FLUX.2 Small Decoder: a faster, drop-in replacement for our standard decoder.
— Black Forest Labs (@bfl_ml) April 8, 2026
→ ~1.4x faster
→ Lower peak VRAM - decode larger images without running out of memory
→ Minimal quality loss
→ Works with FLUX.2 out of the box
Especially impactful for real-time and… pic.twitter.com/AYnonpOgGI
The FLUX.2 Small Decoder is a decoder compatible with the FLUX.2 series of image generation AI, and it can reduce the VRAM usage for decoding to 1.4 times while increasing the decoding speed by 1.4 times. The FLUX.2 Small Decoder can also be used with the open models 'FLUX.2-klein-4B', 'FLUX.2-klein-9B', 'FLUX.2-klein-9b-kv', and 'FLUX.2-dev'.
The FLUX.2 Small Decoder is characterized by its faster processing speed compared to conventional decoders while maintaining the quality of the output image. The image below shows the result of decoding with a conventional decoder on the left and with the FLUX.2 Small Decoder on the right, and they are almost indistinguishable. The decoding time has been reduced from 91.6 milliseconds to 69.4 milliseconds, a 1.32-fold speedup.
Other examples demonstrate that the quality is indistinguishable to the human eye.
The image above shows the difference calculated mechanically, with the differences highlighted in white. You can see that there are slight differences in the outlines and lettering of the signposts that are indistinguishable to the human eye. Black Forest Labs boasts that the quality of their FLUX.2 Small Decoder is 'minimal, almost zero, degradation.'
FLUX.2 Small Decoder can be downloaded from the following link.
black-forest-labs/FLUX.2-small-decoder · Hugging Face
https://huggingface.co/black-forest-labs/FLUX.2-small-decoder
Related Posts:
in AI, Posted by log1o_hf