Gemma 3 (27B)
Google open model with strong multilingual capabilities
Model Summary
Family
Gemma
Version
3.0
Parameters
27B (est.)
Parameter counts for closed models are estimates; vendors rarely publish exact sizes.
VRAM Requirements by Quantization
Memory needed to serve Gemma 3 (27B) for inference, including a 20% overhead for activations and KV cache.
| Precision | VRAM needed | Smallest single GPU that fits |
|---|---|---|
| INT4 (4-bit) | 15.09 GB | NVIDIA P100 SXM2 (16 GB) |
| INT8 (8-bit) | 30.17 GB | NVIDIA V100 SXM2 32GB (32 GB) |
| FP16 (16-bit) | 60.35 GB | NVIDIA H100 SXM5 (80 GB) |
| FP32 (32-bit) | 120.70 GB | AMD Instinct MI250X (128 GB) |
Recommended GPU Configurations
Cheapest on-demand configurations to serve Gemma 3 (27B) at 8-bit (30 GB VRAM).
2x NVIDIA T4
32 GB total VRAM · Turing
~$1.00/h
2x NVIDIA P100 SXM2
32 GB total VRAM · Pascal
~$1.20/h
1x NVIDIA A40
48 GB total VRAM · Ampere
~$1.80/h
Quick GPU Planning
Use the calculator pre-filled with this exact version to estimate memory, speed, and compute requirements in a few clicks.
Access Pre-filled CalculatorFrequently Asked Questions
How much VRAM do you need to run Gemma 3 (27B)?
With an estimated 27B parameters, Gemma 3 (27B) needs roughly 30 GB of VRAM in 8-bit (INT8), 15 GB in 4-bit, and 60 GB in FP16, including a 20% overhead for activations and KV cache.
Which GPUs can run Gemma 3 (27B)?
At 8-bit quantization, the most cost-effective option is 2x NVIDIA T4 (32 GB combined VRAM, around $1.00/hour on-demand). Higher-end cards like the NVIDIA B200 or AMD MI355X reduce the GPU count needed.
Can Gemma 3 (27B) run on a single GPU?
Yes. In 8-bit, a single NVIDIA V100 SXM2 32GB (32 GB) fits the model.
How much does it cost to serve Gemma 3 (27B) in the cloud?
Renting 2x T4 costs on the order of $1.00/hour, i.e. about $730/month running 24/7. Actual prices vary by provider and commitment; spot and reserved capacity can be significantly cheaper.
Deploy on a GPU cloud
Rent 2x T4 by the hour instead of buying hardware.
