Gemma 3 27B: how much VRAM?
In Q4_K_M with 8k tokens of context, Gemma 3 27B needs about 19 GB of memory. The cheapest option: 2 × RTX 3060 12 GB (€1,460 for the complete machine).
Model sheet
ollama run gemma3:27bWeights on Hugging Face: google/gemma-3-27b-it
VRAM by precision and context
| Precision | 4k | 8k | 32k | 128k |
|---|---|---|---|---|
| FP16 | 58 GB | 59 GB | 65 GB | 89 GB |
| FP8 | 30 GB | 31 GB | 37 GB | 61 GB |
| Q8_0 | 31 GB | 32 GB | 38 GB | 62 GB |
| Q6_K | 25 GB | 26 GB | 32 GB | 56 GB |
| Q4_K_M | 18 GB | 19 GB | 25 GB | 49 GB |
| Q3_K_M | 15 GB | 16 GB | 22 GB | 46 GB |
One user. Each simultaneous user adds their own context cache.
Frequently asked questions
How much VRAM does Gemma 3 27B need?
With an 8k-token context and one user, Gemma 3 27B needs about 19 GB in Q4, 32 GB in Q8 and 59 GB in FP16.
What is the cheapest hardware to run Gemma 3 27B?
In Q4, the cheapest option in our catalogue to run Gemma 3 27B is 2 × RTX 3060 12 GB, about €1,460 for the complete machine.
How fast does Gemma 3 27B run?
Expect around 18 tokens per second for one user on 2 × RTX 3060 12 GB in Q4 (estimate based on memory bandwidth).
On which hardware?
| Hardware | Cards | Speed | Machine |
|---|---|---|---|
| RTX 3060 12 GB | 2 cards | ~18 tok/s | €1,460 |
| RTX 3090 24 GB (used) | 1 is enough | ~31 tok/s | €1,650 |
| RTX 4060 Ti 16 GB | 2 cards | ~15 tok/s | €1,800 |
| Radeon RX 7900 XTX 24 GB | 1 is enough | ~32 tok/s | €1,850 |
| Radeon RX 7800 XT 16 GB | 2 cards | ~31 tok/s | €1,900 |
| RTX 4070 12 GB | 2 cards | ~25 tok/s | €2,000 |
| Mini-PC Ryzen AI Max+ 395 128 GB | 1 is enough | ~8.6 tok/s | €2,300 |
| RTX 4070 Ti SUPER 16 GB | 2 cards | ~34 tok/s | €2,600 |
| RTX 4090 24 GB | 1 is enough | ~34 tok/s | €2,900 |
| RTX 4080 SUPER 16 GB | 2 cards | ~37 tok/s | €3,000 |
| RTX 5080 16 GB | 2 cards | ~48 tok/s | €3,300 |
| RTX 5090 32 GB | 1 is enough | ~60 tok/s | €3,500 |
| NVIDIA DGX Spark 128 GB | 1 is enough | ~9.2 tok/s | €4,000 |
| Mac Studio M4 Max 128 GB | 1 is enough | ~18 tok/s | €4,400 |
| RTX 6000 Ada 48 GB | 1 is enough | ~32 tok/s | €7,900 |
| RTX PRO 6000 Blackwell 96 GB | 1 is enough | ~60 tok/s | €9,900 |
| Mac Studio M3 Ultra 512 GB | 1 is enough | ~28 tok/s | €11,500 |
| L40S 48 GB | 1 is enough | ~29 tok/s | €13,500 |
| A100 80 GB | 1 is enough | ~69 tok/s | €17,000 |
| H100 80 GB | 1 is enough | ~113 tok/s | €34,000 |
| H200 141 GB | 1 is enough | ~161 tok/s | €39,000 |
| B200 192 GB | 1 is enough | ~269 tok/s | €48,000 |
Estimated complete machine: cards + host PC (€900 for 1–2 consumer cards, €2,800 beyond, a €6,000 server for datacenter cards). Macs and mini-PCs are sold complete.
Similar models
From: RTX 3060 12 GB · €1,180
From: RTX 4060 Ti 16 GB · €1,350
From: 2 × RTX 3060 12 GB · €1,460
From: 2 × RTX 3060 12 GB · €1,460