MLX bf16 conversion of google/diffusiongemma-26B-A4B-it (mlx-vlm).
Block-diffusion LM built on Gemma 4 (25.2B total / 3.8B active, MoE 128+1 experts, vision).
google/diffusiongemma-26B-A4B-it (license: apache-2.0)mlx_vlm.convert (no quantization (format conversion only)), ~16 bpwmodel_type: diffusion_gemma is supported natively by mlx-vlm 0.6.9; no patch needed.chat_template.jinja is not the base repo's copy: it is the patched Gemma 4 Canonical
Chat Template from ToPo-ToPo/gemma-4-26B-A4B-it-mlx-4bit,
which suppresses the thinking channel when enable_thinking is false (otherwise the literal
word thought leaks into the answer). Thinking is off by default. Weights are unaffected —
restore the base repo's template for stock behaviour.from mlx_vlm import load
model, processor = load("ToPo-ToPo/diffusiongemma-26B-A4B-it-mlx-bf16")
Diffusion generation takes its own flags:
python -m mlx_vlm generate --model ToPo-ToPo/diffusiongemma-26B-A4B-it-mlx-bf16 \
--prompt "Why is the sky blue?" \
--max-tokens 256 --max-denoising-steps 48 --diffusion-sampler entropy-bound
model-00010-of-00011.safetensors 4.9 GBmodel-00004-of-00011.safetensors 4.9 GBmodel-00008-of-00011.safetensors 4.9 GBmodel-00006-of-00011.safetensors 4.9 GBmodel-00002-of-00011.safetensors 4.9 GBmodel-00005-of-00011.safetensors 4.9 GBmodel-00009-of-00011.safetensors 4.9 GBmodel-00007-of-00011.safetensors 4.9 GBmodel-00003-of-00011.safetensors 4.9 GBmodel-00001-of-00011.safetensors 4.8 GBmodel-00011-of-00011.safetensors 2.7 GB有文件？提交种子。
提交种子