hugpadhugpad

AlexHung29629 / gemma-4-26B-A4B-it-dequant-Q4KM

Vision26B280Updated Oct 30, 2026
Hugging Face

Downloads

28

Likes

0

Last updated

Oct 30, 2026

License

apache-2.0

⚠️ This is NOT the original Google checkpoint. These weights were reconstructed by dequantizing a 4-bit GGUF quantization back to bfloat16. All quantization error is baked in — converting to bf16 does not recover the lost precision. Quality is that of the UD-Q4KM quant, not o…

Tasks & library

  • image text to text
  • transformers

Tags

  • transformers
  • gemma4
  • image-text-to-text
  • gemma
  • dequantized
  • gguf-to-safetensors

Technical specifications

Parameters26B
Librarytransformers
LanguagesNot available
Base modelgoogle/gemma-4-26B-A4B-it
Licenseapache-2.0