DI

DiffusionGemma 26B A4B it ROCmFP4

DiffusiongemmaGGUF by kingjones777Updated 2026-08-18
Download on Hugging Face
Parameters25.3B
Context256K
Layers30
Architecturediffusion-gemma
Smallest quant14.4 GB
Quick answer

DiffusionGemma 26B A4B it ROCmFP4 needs about 32.0 GB of memory at Q6_0 (a 23.4 GB file) with an 8K context window, so it fits on a 32 GB GPU. This repository has 4 quants, from 14.4 GB (Q4_0) to 26.5 GB (Q8_0). Longer context needs more: at 32K tokens the KV cache alone adds about 32.2 GB.

DiffusionGemma 26B A4B it ROCmFP4 – a block‑diffusion multimodal model

DiffusionGemma is a 26.3 billion‑parameter MoE model built on the Gemma 4 architecture. It generates text via discrete diffusion, processing text, image, and video inputs to produce text output. The model is optimized for low‑latency, high‑speed inference on a single accelerator.

Good at

  • Fast block‑diffusion decoding
  • Multimodal input handling
  • Reasoning with configurable thinking modes

Limits

  • Text‑only; no vision or video generation
  • Requires 8 k context for memory estimates
  • No mmproj or MTP draft head included

Best for

  • Local inference on AMD RDNA3.5 hardware
  • Low‑latency text generation
  • Reasoning tasks with MoE efficiency

Who it’s for. Users needing a fast, local text generation model that can process interleaved text, image, and video inputs. Ideal for developers deploying on AMD GPUs with limited disk space and seeking efficient reasoning capabilities.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
Q4_014.4 GB 23.0 GB
Too big Get file →
Q6_023.4 GB 32.0 GB
Too big Get file →
Q8_026.1 GB 34.7 GB
Too big Get file →
Q8_026.5 GB 35.1 GB
Too big Get file →

For 8 GB VRAM, the Q4_0 quant (~14.45 GB) is the smallest but requires ~23 GB memory at 8 k context. Q6_0 (~23.43 GB) uses ~32 GB memory. Q8_0 builds (~26 GB) use ~34.7–35.1 GB memory. Choose the smallest quant that fits your VRAM and disk constraints.

What GPU do you need for DiffusionGemma 26B A4B it ROCmFP4?

These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.

  • 4 GB VRAM: no quant fits at an 8K context.
  • 6 GB VRAM: no quant fits at an 8K context.
  • 8 GB VRAM: no quant fits at an 8K context.
  • 12 GB VRAM: no quant fits at an 8K context.
  • 16 GB VRAM: no quant fits at an 8K context.
  • 24 GB VRAM: the best quant that fits is Q4_0 (14.4 GB file, about 23.0 GB in use).
  • 32 GB VRAM: the best quant that fits is Q6_0 (23.4 GB file, about 32.0 GB in use).
  • 48 GB VRAM: the best quant that fits is Q8_0 (26.5 GB file, about 35.1 GB in use).

Which DiffusionGemma 26B A4B it ROCmFP4 quant should you choose?

Q6_0 (23.4 GB) is the middle option in this repository. Very close to the original quality for most uses, at a clearly larger size than 4-bit.

If memory is tight, the smallest file is Q4_0 at 14.4 GB, with the quality trade-offs that come with fewer bits. The usual sweet spot between file size and quality, and the quant most people start with.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architecturediffusion-gemma
Parameters25.3B (approximate, mixture of experts)
Active parameters4B
Layers30
Trained context262,144 tokens
KV cache (fp16)960 KB per token

Source

Quantized bykingjones777
Licensegemma
Downloads (HF)721
Last updated2026-08-18

Frequently asked questions about DiffusionGemma 26B A4B it ROCmFP4

How much VRAM does DiffusionGemma 26B A4B it ROCmFP4 need?

About 32.0 GB at Q6_0 with an 8K context window. That is the 23.4 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.

Can I run DiffusionGemma 26B A4B it ROCmFP4 on an 8 GB GPU?

Not at an 8K context. The smallest file needs about 23.0 GB, so you would need a larger GPU or CPU offloading.

What is the file size of DiffusionGemma 26B A4B it ROCmFP4 GGUF?

It depends on the quant: Q4_0 is 14.4 GB, Q6_0 is 23.4 GB, Q8_0 is 26.1 GB, Q8_0 is 26.5 GB.

What context length does DiffusionGemma 26B A4B it ROCmFP4 support?

The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 8.1 GB at 8K and 32.2 GB at 32K.

Where can I download DiffusionGemma 26B A4B it ROCmFP4 GGUF?

The files are on Hugging Face in the kingjones777/DiffusionGemma-26B-A4B-it-ROCmFP4-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.

What license does DiffusionGemma 26B A4B it ROCmFP4 use?

Hugging Face lists the license as gemma. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.

Can I run DiffusionGemma on 8 GB VRAM?

The smallest quant (Q4_0) requires ~23 GB memory at 8 k context, so 8 GB VRAM is insufficient.

Does this model support image generation?

No, the model is text‑only; it processes image inputs but outputs text.

What hardware is recommended?

AMD RDNA3.5 GPUs such as Strix Halo (gfx1151) are supported; use llama-diffusion-cli or llama-diffusion-gemma-server.

Other GGUFs of diffusiongemma-26B-A4B-it

Similar models

Browse all →

More from kingjones777

All by kingjones777 →

More Diffusiongemma models

All Diffusiongemma →

Other medium models

All medium models →