26B A4B needs about 16.0 GB of memory at Q4_0 (a 14.4 GB file) with an 8K context window, so it fits on a 16 GB GPU. This repository has 6 quants, from 0.5 GB (Q8_0) to 26.5 GB (Q8_0). Longer context needs more: at 32K tokens the KV cache alone adds about 4.3 GB.
Gemma‑4 26B A4B: multimodal LLM with 256K context and 25.2B parameters
Gemma‑4 26B A4B is an instruction‑tuned multimodal model that supports text and image input. It offers a 256K token context window and is built for efficient local execution on AMD GPUs. The model is available in several GGUF quantizations.
Good at
- Multimodal text and image understanding
- Large context reasoning
- Efficient AMD GPU inference
Limits
- Requires ROCmFPX‑capable llama.cpp
- No native vision support in standard builds
- Large VRAM needed for higher quants
Best for
- Local inference on AMD GPUs
- Multimodal chat applications
- Reasoning tasks with long context
Who it’s for. Developers and researchers who need a large multimodal model that runs on AMD hardware. The model is suitable for on‑device inference and edge deployment where a 256K context window is beneficial.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q8_0 | 0.5 GB | 2.0 GB | Fits | Get file → | |
| Q4_0 | 13.7 GB | 15.2 GB | Too big | Get file → | |
| Q4_0 | 14.4 GB | 16.0 GB | Too big | Get file → | |
| Q6_0 | 23.4 GB | 25.0 GB | Too big | Get file → | |
| Q8_0 | 26.1 GB | 27.7 GB | Too big | Get file → | |
| Q8_0 | 26.5 GB | 28.1 GB | Too big | Get file → |
For 8 GB VRAM use Q8_0 (0.46 GB file, ~2 GB RAM). For 16 GB VRAM use Q4_0 (13.67 GB file, ~15.2 GB RAM). For 32 GB VRAM use Q6_0 (23.42 GB file, ~25 GB RAM).
What GPU do you need for 26B A4B?
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: the best quant that fits is Q8_0 (0.5 GB file, about 2.0 GB in use).
- 6 GB VRAM: the best quant that fits is Q8_0 (0.5 GB file, about 2.0 GB in use).
- 8 GB VRAM: the best quant that fits is Q8_0 (0.5 GB file, about 2.0 GB in use).
- 12 GB VRAM: the best quant that fits is Q8_0 (0.5 GB file, about 2.0 GB in use).
- 16 GB VRAM: the best quant that fits is Q4_0 (14.4 GB file, about 16.0 GB in use).
- 24 GB VRAM: the best quant that fits is Q4_0 (14.4 GB file, about 16.0 GB in use).
- 32 GB VRAM: the best quant that fits is Q8_0 (26.5 GB file, about 28.1 GB in use).
Which 26B A4B quant should you choose?
Q4_0 (14.4 GB) is the middle option in this repository. The usual sweet spot between file size and quality, and the quant most people start with.
If memory is tight, the smallest file is Q8_0 at 0.5 GB, with the quality trade-offs that come with fewer bits. Close to lossless. It costs roughly twice the memory of a 4-bit file, so it suits cases where you have plenty of VRAM.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about 26B A4B
How much VRAM does 26B A4B need?
About 16.0 GB at Q4_0 with an 8K context window. That is the 14.4 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 26B A4B on an 8 GB GPU?
Yes, with Q8_0: the 0.5 GB file needs about 2.0 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of 26B A4B GGUF?
It depends on the quant: Q8_0 is 0.5 GB, Q4_0 is 13.7 GB, 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 26B A4B support?
The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 1.1 GB at 8K and 4.3 GB at 32K.
Where can I download 26B A4B GGUF?
The files are on Hugging Face in the kingjones777/Gemma-4-26B-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 26B A4B 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 Gemma‑4 26B A4B on 8 GB VRAM?
Yes, using the Q8_0 quantization which requires about 2 GB of VRAM at 8 k context.
Does this model support vision?
Yes, it can process images, but standard builds need a drafter and MTP configuration.
What llama.cpp build is required?
A ROCmFPX‑capable llama.cpp build is necessary; stock builds will not load the Q4_0_ROCMFP4 or Q*_0_ROCMFPX types.