Gemma 4 26B-A4B
MoEVisionGoogle · 27B (4B active) · Mixture of Experts
Gemma 4 MoE base model (official)
0
0
2026-04
256K tokens
Use cases
Your hardware
16 GB Reference Desktop
Quantization options
What is quantization?| Quant | Bits | VRAM | Disk | Quality | Status |
|---|---|---|---|---|---|
| Q2_K | 2 | 9.1 GB | 8.3 GB | low | Decent · ~22 t/s |
| Q3_K_M | 3 | 12.6 GB | 11.6 GB | moderate | Tight fit · ~16 t/s |
| Q4_K_M | 4 | 14.3 GB | 13.2 GB | good | Barely runs · ~14 t/s |
| Q5_K_M | 5 | 17.8 GB | 16.5 GB | good | Barely runs · ~6 t/s |
| Q6_K | 6 | 21.2 GB | 19.8 GB | excellent | Barely runs · ~4 t/s |
| Q8_0 | 8 | 28.2 GB | 26.4 GB | excellent | Barely runs · ~2 t/s |
| F16 | 16 | 55.8 GB | 52.8 GB | lossless | Too heavy · ~1 t/s |
About this model
Gemma 4 26B-A4B (Gemma 4 26b a4b) is a 27B sparse mixture-of-experts (MoE) model activating 4B active per token from Google in the Gemma family, designed primarily for vision, reasoning workflows. Supports up to 256K context tokens; start at 4K–8K with Q8_0 KV cache on tight GPUs.
Hardware planning baselines start at 15.1 GB minimum memory and 25.1 GB recommended memory. At Q4_K_M, weights plus baseline runtime overhead use roughly 14.3 GB of RAM/VRAM and 13.2 GB of disk space. Because MoE routing evaluates only a subset of experts on each generated token, Gemma 4 26B-A4B delivers faster decoding throughput per gigabyte of memory bandwidth than a dense checkpoint of identical total size—provided the full 14.3 GB checkpoint stays resident without SSD paging.
Deploy Gemma 4 26B-A4B locally via llama.cpp with Flash Attention enabled. Under Gemma Terms of Use, review official weight cards before commercial redistribution.
Highlights
- Local Fit Baseline15.1 GB minimum and 25.1 GB recommended RAM/VRAM.
- Primary QuantQ4_K_M targets 14.3 GB memory and 13.2 GB disk.
- Context & Capabilities256K max context · Sparse MoE.
System requirements by device
Memory planning targets 15.1 GB minimum RAM/VRAM and 25.1 GB recommended memory. The baseline quant is Q4_K_M, estimated at 14.3 GB memory and 13.2 GB disk. Actual runtime usage shifts with context length, KV cache precision, and OS background overhead.
| Device | Notes |
|---|---|
| Mac (Apple silicon) | Use unified memory headroom (~75% physical RAM max) as the primary fit signal. Start with Q4_K_M and close heavy browser tabs first. |
| Windows / Linux GPU | Keep weights plus KV cache inside dedicated VRAM so Windows WDDM does not spill layers over PCIe into Shared GPU Memory. |
| Phones & Compact Edge Boards | Not practical on phones (15.1 GB min RAM exceeds Android LMK and iOS Jetsam limits). Pick a 2B–4B model on mobile. |
Start with a short session and leave memory for the operating system. A file fitting on disk does not guarantee that it will load in memory.
Estimated memory by quantization
These are rough estimates, not tested minimum requirements. Check the actual file and runtime documentation; context, offloading and operating-system overhead can change the memory needed.
| Quant Tier | Estimated memory | Disk Size | Quality |
|---|---|---|---|
| Q2_K | 9.1 GB | 8.3 GB | low |
| Q3_K_M | 12.6 GB | 11.6 GB | moderate |
| Q4_K_M | 14.3 GB | 13.2 GB | good |
| Q5_K_M | 17.8 GB | 16.5 GB | good |
| Q6_K | 21.2 GB | 19.8 GB | excellent |
| Q8_0 | 28.2 GB | 26.4 GB | excellent |
| F16 | 55.8 GB | 52.8 GB | lossless |
How to run Gemma 4 26B-A4B on your device
Start with a short prompt and the smallest quant that still gives the quality you need. A clean first run is more useful than loading a larger file with no memory headroom.
Run Gemma 4 26B-A4B on Mac
- Install LM Studio or llama.cpp on macOS Apple Silicon.
- Pull Q4_K_M (14.3 GB) first and confirm it stays below ~75% of your Mac's unified memory.
- Enable Flash Attention and cap initial context at 8K until memory pressure stays below 25.1 GB.
- Consult the Mac guide for Activity Monitor tuning and Metal offload tips.
Run Gemma 4 26B-A4B on Windows
- Install LM Studio for Windows for visual VRAM inspection, or Ollama when you want a localhost API endpoint.
- Load Q4_K_M (~14.3 GB) and verify Dedicated GPU Memory stays within VRAM without spilling into Shared GPU Memory.
- Reserve at least 15.1 GB memory minimum and target 25.1 GB for comfortable context growth.
- Enable Q8_0 KV cache or drop one quant tier if generation slows down over PCIe.
Gemma 4 26B-A4B vs alternatives
| Model | Params | Context | Min RAM | License |
|---|---|---|---|---|
| Gemma 4 26B-A4B | 27B (4B active) | 256K | 15.1 GB | Gemma |
| Gemma 2 27B | 27B | 8K | 15.1 GB | Gemma |
| Gemma 3 27B | 27B | 128K | 15.1 GB | Gemma |
| Gemma 4 26B-A4B IT | 27B (4B active) | 256K | 15.1 GB | Gemma |
| DiffusionGemma 26B-A4B IT | 26B (4B active) | 256K | 14.5 GB | Apache 2.0 |
| Gemma 4 31B IT | 33B | 256K | 18.4 GB | Gemma |
Use this table to choose the smallest nearby model that still matches the job. A related model with lower RAM can be the better local choice when the current page fits only at the edge of available memory.
Frequently asked questions
- What is Gemma 4 26B-A4B?
- Gemma 4 26B-A4B is a 27B sparse mixture-of-experts model (4B active active per token) from Google, supporting up to 256K tokens with 15.1 GB minimum RAM and 25.1 GB recommended RAM.
- How do I download and install Gemma 4 26b a4b?
- Download a GGUF build for Gemma 4 26B-A4B, start with Q4_K_M, and confirm the model fits locally before increasing context or quant quality.
- How much VRAM or RAM does Gemma 4 26B-A4B require?
- At Q4_K_M, Gemma 4 26B-A4B occupies roughly 14.3 GB of active memory and 13.2 GB on disk. Add 2–6 GB of spare headroom for KV cache buffers, runtime scratch space, and OS apps (25.1 GB recommended).
- What is the best way to run Gemma 4 26B-A4B locally?
- Run Gemma 4 26B-A4B in llama.cpp starting with Q4_K_M. Keep initial context at 4K–8K tokens and confirm 100% GPU/Metal offload before scaling context length.
- Can Gemma 4 26B-A4B run on Android or iPhone?
- Gemma 4 26B-A4B requires 15.1 GB minimum memory, placing it outside practical 8 GB smartphone limits. Run it on a 16 GB+ Mac or desktop GPU instead, and pick a 2B–4B checkpoint for phones.
- How does Gemma 4 26B-A4B compare to Gemma 2 27B?
- Gemma 2 27B (27B, 15.1 GB min RAM) sits closest in size. Compare active parameters, context ceiling (256K vs 8K), and license terms (Gemma) before switching.