Skip to content

Gemma 4 26B A4B on 2× RTX 5090: VRAM, speed and fit

Gemma 4 26B A4B at 8-bit runs on the 2× RTX 5090 with 256K of context at about 199 tokens a second.

Text generation · Chat · Ollama / LM Studio

Runs comfortably · on 2× RTX 5090 · 8-bit (Q8_0) · Ollama / LM Studio · Chat

GPU memory
30.0 GB
of 64.0 GB
Context
256K
auto
Speed
199 tok/s
137 when full
First reply
0.1 s
2K prompt

Where the memory goes

Model
25.5 GB
Context
2.7 GB
Overhead
1.8 GB
OS
0.0 GB
Free
34.0 GB
Memory
30.0 GB of 64.0 GB
Context
256K tokens · auto
Speed
199 tok/s · 137 when full
First reply
0.1 s for a 2K prompt
Running cost
$17/mo at home · $58/mo on a cloud 2× RTX 2000 Ada

Validation: Validated

What Gemma 4 26B A4B on 2× RTX 5090 was checked against
CheckTargetResultStatus
Fits in GPU memory64.0 GB usable30.0 GB neededPass
Headroom5% free34.0 GB free (53%)Pass
Context for chat16K or more256KPass
Writing speed15 tok/s or more199 tok/sPass
First replyunder 3 s, 2K prompt0.1 sPass

Targets: a reply you can read along with, a first reply the use case tolerates, and 5% of memory left free. Calculated from published specs, not benchmarked.

The setup

Model
Gemma 4 26B A4B
Parameters
25.8B, 3.8B active
Precision
8-bit (Q8_0)
Context
256K, auto
Chats at once
1
Use
Chat
Runs in
Ollama / LM Studio
Hardware
2× RTX 5090
GPU can use
64.0 GB of 64.0 GB

Running cost

At home
$17/mo
770 W at the wall
Renting instead
$58/mo
2× RTX 2000 Ada at $0.48/h

4 h a day, power at $0.18 per kWh, cloud at RunPod on-demand prices as of Nov 22, 2026.

How to run it

Ollama
ollama run gemma4:26b
llama.cpp
llama-server -hf unsloth/gemma-4-26B-A4B-it-GGUF
vLLM
vllm serve google/gemma-4-26B-A4B-it
MLX (Mac)
mlx_lm.chat --model mlx-community/gemma-4-26b-a4b-it-4bit

Gemma 4 26B A4B on similar machines

All machines: Gemma 4 26B A4B VRAM requirements. Everything that runs here: what can the 2× RTX 5090 run?

Other models that run on 2× RTX 5090

Questions

Can Gemma 4 26B A4B run on 2× RTX 5090?

Gemma 4 26B A4B at 8-bit runs on the 2× RTX 5090 with 256K of context at about 199 tokens a second.

Report VR-13Y9D0M, calculated Nov 22, 2026 from published specs, not benchmarked.