Gemma 4 GPU Requirements: Every Variant (2026)

Gemma 4 is a family, not a model, and its VRAM requirements span a factor of ten. The E2B edge variant loads in 2.3 GB while the 31B dense flagship wants 20 GB. Picking the wrong variant for your hardware is the most common Gemma 4 mistake, and we see it constantly in sizing questions. We ran all five variants through the ModelFit engine on September 3, 2026, and this guide maps each one to the GPUs and Macs that actually run it, with engine-estimated speeds.

TL;DR: E2B (2.3 GB) runs on anything including phones. E4B (4 GB) is the 8GB-GPU pick. 12B (8 GB) is the sweet spot for 12-16GB GPUs and 16GB Macs. 26B-A4B (16 GB, MoE) wants 24GB. 31B (20 GB, dense) is the quality flagship for 24-32GB. Every variant except E2B is a strong local model in 2026.

How much VRAM does each Gemma 4 variant need?

All sizes are Q4_K_M as shipped via Ollama, consistent with the builds on Google's Hugging Face org and the Gemma documentation. These are the builds ModelFit tracks, and you can line them up against other families on our quant comparison tool.

Context adds on top of every figure in that table, and it is the part most sizing guides leave out. The 12B and 31B are dense GQA models with standard KV cache costs of roughly 128-192 KB per token at fp16, so a long document session can quietly double what the weights alone suggested. The 26B-A4B, in contrast, is multimodal, and its mixture-of-experts design decodes at the speed of a 4B model while occupying 16 GB, which is why its row in the speed tables below looks like a typo until you check the architecture.

Which GPUs fit each variant?

Speeds are ModelFit engine estimates at Q4_K_M, labeled est.

Gemma 4 12B (the most searched)

On Macs, the 12B is the flagship pick for 16GB machines: ~19 tok/s est. on a Mac Mini M6 16GB, ~17 tok/s est. on a fanless MacBook Air M5 16GB. For example, a full coding-assistant session at 16K context stays responsive on both, though the Air will give a little of that speed back once its chassis warms up under a long generation.

Gemma 4 26B-A4B and 31B

The 26B-A4B is the faster of the two on the same card because its 4B active path reads far fewer bytes per token, and at this memory tier bandwidth is the binding constraint. On 24GB it runs about twice as fast as the dense 31B while scoring within a point of it on quality benchmarks, which makes the choice look strange until you realize what the 31B is for: maximum dense quality for buyers who specifically distrust MoE routing. For most buyers, the MoE is the right 24GB pick — we measured the gap on every 24GB card in our catalog and it holds everywhere. Per-model detail: 26B-A4B model page, 31B model page.

Which Variant Should You Pick?

• 8GB GPU (RTX 4060, 3070): E4B at ~49 tok/s est., or Qwen 3.5 4B if you want a sharper small model. The 12B does not fit, however.
• 12GB GPU (RTX 3060, 5070): 12B is the headline. ~30-42 tok/s est. See the Gemma 4 12B model page.
• 16GB GPU (RTX 5060 Ti, 5080): 12B with room for context, or 26B-A4B tight. Prefer 24GB for the MoE.
• 24GB GPU (RTX 3090, 4090): 26B-A4B for speed or 31B for maximum dense quality. Hardware-by-hardware detail is on the 26B-A4B best-hardware page.
• 16GB Mac: 12B at ~17-19 tok/s est. is the best current-gen fit.
• 24GB Mac: 26B-A4B at ~21 tok/s est. on an M6 Mac Mini.
• 48GB+ Mac: 31B at ~13 tok/s est. on an M5 Pro Mac Mini 48GB, or step up to Qwen 3.6 35B-A3B.

How do the E2B and E4B edge variants differ in practice?

The two small variants get lumped together because both are tiny, but they are built for different jobs and the difference matters once you pick hardware around them. E2B at 2.3 GB is an edge model in the literal sense — phones, embedded boards, and browser demos — and it makes compromises on reasoning that you will notice within a few prompts. E4B at 4 GB is the smallest Gemma 4 that behaves like a real assistant, and on an 8GB GPU it leaves enough headroom for context that it works as a daily driver rather than a demo.

The mistake we see is buying for E2B's footprint and then expecting E4B's answers. If your workload lives on a phone or a microcontroller, E2B is the honest choice; if it lives on a desktop GPU with 8GB, E4B is the floor we would recommend, and the jump to the 12B is worth more than either.

FAQ

Is Gemma 4 12B enough for coding?

Yes. It scores well on LiveCodeBench and is our top pick for a 16GB Mac. For dedicated coding agents, Ornith 1.0 9B is faster at the same memory footprint, but Gemma 4 12B is the better general model.

What is the difference between 26B-A4B and 31B?

Architecture. The 31B is a traditional dense model. The 26B-A4B is a mixture of experts with 4B active per token, so it runs roughly twice as fast on the same hardware with nearly the same quality. Unless you specifically want dense weights, the MoE is the better 24GB choice.

Does Gemma 4 E2B run on a phone?

Yes. At 2.3 GB it is designed for on-device use and runs on recent iPhones and Android flagships. It is the weakest Gemma 4 by a wide margin, so treat it as an edge model, not a desktop assistant.

Why does the 12B not fit on an 8GB GPU?

The weights are 8 GB before the OS, display server, and context cache. On an 8GB card the model spills into shared system memory and collapses to single-digit tokens per second. On Windows the problem is worse because the OS reserves more VRAM. Use E4B on 8GB.

___

All five variant tables here come from the ModelFit engine, re-run on September 3, 2026, with speeds labeled est. How we work: about ModelFit. Found a stale number? Tell us.