Only with CPU offload

Gemma 4 26B-A4B at Q4_K_M needs 16.0 GB but only 14.4 GB is addressable, so about 11% of the layers would stream from system RAM at roughly 60 GB/s. Expect around 36 tokens per second — usable for batch work, painful for chat.

89% on GPU 8K context Q4_K_M · 14.9 GB Apache 2.0 Released 2 Apr 2026 Vision

Mixture of experts with 3.8B active. Slower to think than Qwen3.6 35B-A3B, faster to answer, and it sees images.

Hardware requirements What hardware do I need to run Gemma 4 26B-A4B? →

At Q4_K_M with an 8K context: a 24 GB GPU (RTX 4090) or a 32 GB Mac (M4). Minimum, recommended, high-quality and long-context tiers, with all 114 devices checked at every quantisation.

Fits instead: Gemma 4 12B (8.2 GB) · Gemma 4 E4B (5.2 GB)

The VRAM budget

weights 14.9 GB
Weights 14.9 GB KV cache @ 8K 0.51 GB Runtime overhead 0.6 GB Over budget 1.6 GB past 14.4 GB

Quantisation ladder

QuantWeightsTotal @ 8KMax contextTok/sQualityFit
Q8_0 26.2 GB 27.3 GB — ~6.9 −0.1% ppl 12.9 GB over
Q6_K 20.2 GB 21.3 GB — ~12 −0.4% ppl 6.9 GB over
Q5_K_M 17.5 GB 18.6 GB — ~18 −0.8% ppl 4.2 GB over
Q4_K_M 14.9 GB 16.0 GB — ~36 −1.9% ppl 1.6 GB over
Q3_K_M 12.1 GB 13.2 GB 39K 99 −5.4% ppl Long context
Q2_K 10.3 GB 11.4 GB 83K 115 −15% ppl Long context

Quality is the published perplexity delta against f16 weights. Max context assumes an f16 KV cache; q8_0 roughly doubles it. This model interleaves sliding-window layers (1024 tokens, 5 of 30 layers global), which is why its cache barely grows with context.

How to run it

terminal
$ llama-server \
    -hf google/gemma-4-26B-A4B-it:Q4_K_M \
    -c 8192 -ngl 26

The engine underneath most of the others. Every knob is exposed. More on llama.cpp.

01Download is 14.9 GB. Keep it on an SSD — a first load off a spinning disk takes minutes.
02Close anything else holding VRAM. A browser with hardware acceleration can sit on 1–2 GB.
03If it falls back to CPU silently, drop the context first, then step down a quantisation.
See all models for this rig Compare with another model

Related models on a Tesla P100 16 GB

ModelTotalTok/s est.
Gemma 4 31B
31.3B · same family
Q4_K_M · 20.2 GB ~5.3 CPU offload
Gemma 4 12B
12B · same family
Q4_K_M · 8.2 GB 66 Runs great
Gemma 4 E4B
8.0B · same family
Q4_K_M · 5.2 GB 99 Runs great
Gemma 4 E2B
5.1B · same family
Q4_K_M · 3.5 GB 155 Runs great
Gemma 3 27B
27.4B · similar size
Q4_K_M · 17.0 GB ~9.9 CPU offload
Qwen3.8 27B
27.8B · similar size
Q4_K_M · 16.7 GB ~11 CPU offload
Qwen3.6 27B
27.8B · similar size
Q4_K_M · 16.7 GB ~11 CPU offload

Every model against the Tesla P100 16 GB →

Gemma 4 26B-A4B on other hardware

DeviceTotalTok/s est.
GeForce RTX 4060 Laptop
7.0 GB usable · 256 GB/s
16.0 GB ~9.4 CPU offload
GeForce RTX 3060 12 GB
10.6 GB usable · 360 GB/s
16.0 GB ~14 CPU offload
M4 · 16 GB
10.7 GB usable · 120 GB/s
16.0 GB — Won't fit
GeForce RTX 4090
22.4 GB usable · 1008 GB/s
16.0 GB 110 Runs great
Radeon RX 7900 XTX
22.4 GB usable · 960 GB/s
16.0 GB 105 Runs great
M4 · 32 GB
24.0 GB usable · 120 GB/s
16.0 GB 15 Runs great
M3 Max · 64 GB
48.0 GB usable · 400 GB/s
16.0 GB 49 Runs great

All 114 devices, by tier →