Skip to content

Gemma 4 26B A4B VRAM requirements

How much VRAM Gemma 4 26B A4B needs
4-bit weights: 14.6 GBWith 8K context: 15.9 GBWith 32K context: 16.1 GB8-bit, 8K context: 26.8 GB
Parameters
25.8B
Active per token
3.8B
Longest context
256K
Released
Mar 2026
License
Apache 2.0
Strongest at
Chat, Coding, Agents, Docs

How much VRAM Gemma 4 26B A4B needs

Memory Gemma 4 26B A4B needs at each precision and context length
PrecisionWeightsWith 8KWith 32KWith 128K
16-bit48.1 GB49.4 GB49.6 GB50.6 GB
8-bit25.5 GB26.8 GB27.1 GB28.0 GB
6-bit19.7 GB21.0 GB21.3 GB22.2 GB
5-bit17.1 GB18.4 GB18.6 GB19.6 GB
4-bit14.6 GB15.9 GB16.1 GB17.1 GB
3-bit11.7 GB13.1 GB13.3 GB14.2 GB
2-bit9.0 GB10.3 GB10.6 GB11.5 GB
1.6-bit5.0 GB6.3 GB6.5 GB7.4 GB

One conversation, 16-bit context cache, runtime overhead included.

GPUs and Macs that run Gemma 4 26B A4B

The best way Gemma 4 26B A4B runs on common GPUs, Macs and AI boxes
MachineBest fitContextSpeedFirst replyReport
RTX 4060Too bigPDF: Gemma 4 26B A4B on RTX 4060 report
RTX 5060Too bigPDF: Gemma 4 26B A4B on RTX 5060 report
RTX 3070Too bigPDF: Gemma 4 26B A4B on RTX 3070 report
RTX 3080 10 GBToo bigPDF: Gemma 4 26B A4B on RTX 3080 10 GB report
RTX 3060 12 GBToo bigPDF: Gemma 4 26B A4B on RTX 3060 12 GB report
Arc B580Too bigPDF: Gemma 4 26B A4B on Arc B580 report
RTX 4070Too bigPDF: Gemma 4 26B A4B on RTX 4070 report
RTX 4070 SuperToo bigPDF: Gemma 4 26B A4B on RTX 4070 Super report
RTX 5070Too bigPDF: Gemma 4 26B A4B on RTX 5070 report
RTX 4060 Ti 16 GB3-bit224K91 tok/s1.1 sPDF: Gemma 4 26B A4B on RTX 4060 Ti 16 GB report
RTX 5060 Ti 16 GB3-bit224K129 tok/s1.0 sPDF: Gemma 4 26B A4B on RTX 5060 Ti 16 GB report
Arc A770 16 GB3-bit224K104 tok/s1.1 sPDF: Gemma 4 26B A4B on Arc A770 16 GB report
Radeon RX 9070 XT3-bit224K133 tok/s0.6 sPDF: Gemma 4 26B A4B on Radeon RX 9070 XT report
RTX 4070 Ti Super3-bit224K174 tok/s0.5 sPDF: Gemma 4 26B A4B on RTX 4070 Ti Super report
RTX 4080 Super3-bit224K185 tok/s0.4 sPDF: Gemma 4 26B A4B on RTX 4080 Super report
RTX 5070 Ti3-bit224K210 tok/s0.5 sPDF: Gemma 4 26B A4B on RTX 5070 Ti report
RTX 50803-bit224K219 tok/s0.4 sPDF: Gemma 4 26B A4B on RTX 5080 report
Radeon RX 7900 XTX6-bit184K123 tok/s1.0 sPDF: Gemma 4 26B A4B on Radeon RX 7900 XTX report
RTX 30906-bit184K153 tok/s0.7 sPDF: Gemma 4 26B A4B on RTX 3090 report
RTX 40906-bit184K161 tok/s0.3 sPDF: Gemma 4 26B A4B on RTX 4090 report
Radeon AI PRO R97008-bit256K76 tok/s0.7 sPDF: Gemma 4 26B A4B on Radeon AI PRO R9700 report
RTX 50908-bit256K199 tok/s0.2 sPDF: Gemma 4 26B A4B on RTX 5090 report
2× RTX 30908-bit256K126 tok/s0.4 sPDF: Gemma 4 26B A4B on 2× RTX 3090 report
2× RTX 40908-bit256K133 tok/s0.2 sPDF: Gemma 4 26B A4B on 2× RTX 4090 report
RTX PRO 5000 Blackwell 48 GB8-bit256K164 tok/s0.3 sPDF: Gemma 4 26B A4B on RTX PRO 5000 Blackwell 48 GB report
RTX 6000 Ada8-bit256K128 tok/s0.2 sPDF: Gemma 4 26B A4B on RTX 6000 Ada report
L40S8-bit256K118 tok/s0.4 sPDF: Gemma 4 26B A4B on L40S report
2× RTX 50908-bit256K199 tok/s0.1 sPDF: Gemma 4 26B A4B on 2× RTX 5090 report
4× RTX 30908-bit256K126 tok/s0.2 sPDF: Gemma 4 26B A4B on 4× RTX 3090 report
RTX PRO 6000 Blackwell8-bit256K199 tok/s0.2 sPDF: Gemma 4 26B A4B on RTX PRO 6000 Blackwell report
Mac mini M4 16 GBToo bigPDF: Gemma 4 26B A4B on Mac mini M4 16 GB report
Mac mini M4 24 GB3-bit240K43 tok/s15 sPDF: Gemma 4 26B A4B on Mac mini M4 24 GB report
Mac mini M4 32 GB5-bit216K32 tok/s15 sPDF: Gemma 4 26B A4B on Mac mini M4 32 GB report
Mac mini M4 Pro 24 GB3-bit240K80 tok/s7.4 sPDF: Gemma 4 26B A4B on Mac mini M4 Pro 24 GB report
Mac mini M4 Pro 48 GB8-bit256K45 tok/s7.4 sPDF: Gemma 4 26B A4B on Mac mini M4 Pro 48 GB report
Mac mini M4 Pro 64 GB8-bit256K45 tok/s7.4 sPDF: Gemma 4 26B A4B on Mac mini M4 Pro 64 GB report
M4 Max 128 GB8-bit256K71 tok/s3.7 sPDF: Gemma 4 26B A4B on M4 Max 128 GB report
M5 Max 128 GB8-bit256K77 tok/s1.1 sPDF: Gemma 4 26B A4B on M5 Max 128 GB report
M3 Ultra 96 GB8-bit256K85 tok/s2.1 sPDF: Gemma 4 26B A4B on M3 Ultra 96 GB report
M3 Ultra 256 GB8-bit256K85 tok/s2.1 sPDF: Gemma 4 26B A4B on M3 Ultra 256 GB report
M3 Ultra 512 GB8-bit256K85 tok/s2.1 sPDF: Gemma 4 26B A4B on M3 Ultra 512 GB report
M5 Ultra 512 GB8-bit256K107 tok/s0.6 sPDF: Gemma 4 26B A4B on M5 Ultra 512 GB report
Ryzen AI Max+ 395 64 GB8-bit256K34 tok/s2.1 sPDF: Gemma 4 26B A4B on Ryzen AI Max+ 395 64 GB report
Ryzen AI Max+ 395 128 GB8-bit256K34 tok/s2.1 sPDF: Gemma 4 26B A4B on Ryzen AI Max+ 395 128 GB report
RTX Spark8-bit256K48 tok/s0.7 sPDF: Gemma 4 26B A4B on RTX Spark report
DGX Spark8-bit256K44 tok/s0.7 sPDF: Gemma 4 26B A4B on DGX Spark report
Dual DGX Spark8-bit256K44 tok/s0.4 sPDF: Gemma 4 26B A4B on Dual DGX Spark report
DGX Station GB3008-bit256K383 tok/s0.1 sPDF: Gemma 4 26B A4B on DGX Station GB300 report
2× RTX PRO 6000 Blackwell8-bit256K199 tok/s0.1 sPDF: Gemma 4 26B A4B on 2× RTX PRO 6000 Blackwell report
4× RTX PRO 6000 Blackwell8-bit256K199 tok/s0.1 sPDF: Gemma 4 26B A4B on 4× RTX PRO 6000 Blackwell report
8× RTX PRO 6000 Blackwell8-bit256K199 tok/s0.1 sPDF: Gemma 4 26B A4B on 8× RTX PRO 6000 Blackwell report
A100 80 GB SXM8-bit256K216 tok/s0.5 sPDF: Gemma 4 26B A4B on A100 80 GB SXM report
H100 SXM8-bit256K284 tok/s0.1 sPDF: Gemma 4 26B A4B on H100 SXM report
4× H200 NVL8-bit256K333 tok/s0.1 sPDF: Gemma 4 26B A4B on 4× H200 NVL report
8× H200 NVL8-bit256K333 tok/s0.1 sPDF: Gemma 4 26B A4B on 8× H200 NVL report

Best fit is the highest quality, up to 8-bit, that leaves 16K of context. First reply is the wait for a 2K prompt. Speeds are rough, plus or minus 25%.

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

Questions

How much VRAM does Gemma 4 26B A4B need?

Gemma 4 26B A4B needs 14.6 GB for its weights at 4-bit, 15.9 GB with an 8K context and 16.1 GB with 32K. At 8-bit, 26.8 GB.

What is the minimum VRAM for Gemma 4 26B A4B?

About 13.1 GB at 3-bit with an 8K context, the lowest precision we recommend for a model this size. The smallest machine here that runs it without spilling is the RTX 4060 Ti 16 GB.

How big is Gemma 4 26B A4B?

25.8B parameters (3.8B active per token): 48.1 GB of weights at 16-bit, 14.6 GB at 4-bit.

Can Gemma 4 26B A4B run on an RTX 4090?

Yes. At 6-bit it holds 184K of context and writes about 161 tokens a second.

More models