Skip to content
localmodel.run

Text model · Gemma

Gemma 4 26B-A4B: RAM and VRAM requirements

Gemma 4 26B-A4B needs about 19 GB to run at Q4_K_M (the Q4_K_M GGUF file is ~17.04 GB to download; KV cache and overhead add the rest), or about 28.9 GB at Q8_0. The lightest hardware that runs it is Nvidia GeForce RTX 4090 (24GB).

Gemma family · 26.5B params (Mixture-of-Experts: activates only 3.8B of 26.5B params per token, so generation is faster than the total size suggests) · released Mar 2026.

Shopping for hardware? See what runs Gemma 4 26B-A4B →

License Apache-2.0 · Commercial OK ↓ 11.9M/mo ♥ 1.3K on HuggingFace
Q4_K_M GGUF
17.04 GB
Q8_0 GGUF
26.86 GB
Memory @ Q4 (4k)
~19 GB
Context
256 k

Q4_K_M and Q8_0 are GGUF download (file) sizes. Memory @ Q4 includes the KV cache and overhead, so it runs larger than the file on disk.

Will it run on your device?

Gemma 4 26B-A4B runs on 13 of 40 tracked devices at Q4_K_M.

12 run well 1 tight fit 27 too small
Fit check Q4_K_M
No, not enough memory
needs 19 GB usable 10.5 GB
See the full breakdown

Memory breakdown

Weights (Q4_K_M)17.04 GB
+
KV cache (4k)1.2 GB
+
Overhead0.8 GB
=
Total19 GB

How context length changes it

4k context ~19 GB 32k context ~27.7 GB 128k context ~57.3 GB

Longer context grows the KV cache, which for Gemma 4 26B-A4B is sized by its 3.8B active params, not the full 26.5B. It needs ~19 GB at 4k and ~57.3 GB at 128k.

Quantization sizes

GGUF quantson disk
Quantization Size on disk
Q2_K 11.1 GB est
Q3_K_M 13 GB est
Q4_K_M (default) 17.04 GB
Q5_K_M 18.9 GB est
Q6_K 21.7 GB est
Q8_0 26.86 GB
FP16 50.51 GB

Lower quant = smaller and faster, slightly lower quality. Q4_K_M is the common default.

Run it

Ollama
$ ollama run gemma4:26b-a4b
llama.cpp
$ llama-cli -hf bartowski/google_gemma-4-26B-A4B-it-GGUF:Q4_K_M
LM Studio
$ lms get bartowski/google_gemma-4-26B-A4B-it-GGUF

Which devices can run Gemma 4 26B-A4B?

Apple Silicon Macs

RAM-only laptops

iPhone & iPad

Android

NVIDIA GPUs

AMD GPUs

Same job, different size

Pick by what fits your memory: step down to free up VRAM, or step up if you have headroom.

Next size down LFM2 24B-A2B 24B Next size up Gemma 2 27B 27B

Similar models

Head-to-head

FAQ

How much VRAM or RAM does Gemma 4 26B-A4B need?

At Q4_K_M, Gemma 4 26B-A4B needs about 19 GB (weights ~17.04 GB + KV cache + overhead) at a 4k context. At Q8_0 budget ~28.9 GB.

What is the Q4_K_M GGUF file size of Gemma 4 26B-A4B?

The Q4_K_M GGUF file is about 17.04 GB to download, and the Q8_0 GGUF is about 26.86 GB. That is the weights file on disk; to run it you also need room for the KV cache and overhead, so budget ~19 GB of memory at Q4_K_M.

Can Gemma 4 26B-A4B run on a laptop?

Gemma 4 26B-A4B is large; you need a 24 GB+ GPU or a 32-48 GB Mac at Q4_K_M.

Is Gemma 4 26B-A4B cheaper to run because it is a MoE model?

It is faster, not lighter. Gemma 4 26B-A4B activates only 3.8B of 26.5B params per token (so it runs quickly), but all experts must stay in memory, so it still needs memory for the full 26.5B.

Can I use Gemma 4 26B-A4B commercially?

Yes. Gemma 4 26B-A4B is licensed Apache-2.0, which permits commercial use.

Understand the numbers

Short guides to the ideas behind Gemma 4 26B-A4B's memory and quant figures.

Google's first MoE Gemma: 25.2B total per the model card (26.5B in safetensors incl. vision tower), 3.8B active. Q4_K_M from bartowski, Q8_0/BF16 from unsloth. Multimodal. 256K context. Despite the 17GB Q4 file, inference speed tracks the 3.8B active params.

Sources

Last validated 2026-08-03. Memory figures are estimates. See methodology.