# Gemma 3 27B IT

> OnPrem World · https://onpremworld.com/en/models/gemma-3-27b · Last updated: 2026-10-16 · CC BY 4.0

Google's 27B with image input; 52 of its 62 layers use a 1,024-token sliding window, which keeps the KV cache small. It fits a 20–24 GB card at 4-bit.

|  |  |
|---|---|
| Developer | Google DeepMind |
| Parameters | 27B |
| Native context | 131,072 tokens |
| Licence | Gemma Terms of Use |
| Commercial use | Yes |
| KV cache | 10 global + 52 sliding-window layers (1024 tokens) |
| Released | 2025-03 |

```
ollama run gemma3:27b
```

## Memory needed

Weights + KV cache + runtime overhead, in GiB. Lower quantization = less memory, slightly lower quality.

| Quantization | 4K | 8K | 32K |
|---|---|---|---|
| Q4_K_M | 17.2 GiB | 17.5 GiB | 19.4 GiB |
| Q5_K_M | 20.0 GiB | 20.3 GiB | 22.2 GiB |
| Q6_K | 23.0 GiB | 23.3 GiB | 25.2 GiB |
| Q8_0 | 29.3 GiB | 29.6 GiB | 31.5 GiB |
| FP16 | 54.0 GiB | 54.3 GiB | 56.2 GiB |

## Which hardware runs it (8K)

| Hardware | Memory | Best quant | tok/s (est.) |
|---|---|---|---|
| DGX Station (GB300) | 252 GB | FP16 | 78 |
| GeForce RTX 5090 | 32 GB | Q8_0 | 40 |
| Radeon RX 7900 XT | 20 GB | Q4_K_M | 32 |
| RTX PRO 5000 48GB Blackwell | 48 GB | Q8_0 | 31 |
| GeForce RTX 4090 | 24 GB | Q6_K | 30 |
| Radeon RX 7900 XTX | 24 GB | Q6_K | 28 |
| GeForce RTX 3090 | 24 GB | Q6_K | 28 |
| RTX 6000 Ada Generation | 48 GB | Q8_0 | 22 |
| RTX PRO 6000 Blackwell Workstation Edition | 96 GB | FP16 | 22 |
| RTX PRO 6000 Blackwell Max-Q Workstation Edition | 96 GB | FP16 | 22 |
| RTX PRO 4500 Blackwell | 32 GB | Q8_0 | 21 |
| RTX PRO 4000 Blackwell | 24 GB | Q6_K | 20 |
| RTX A6000 | 48 GB | Q8_0 | 18 |
| RTX PRO 5000 72GB Blackwell | 72 GB | FP16 | 17 |
| Radeon AI PRO R9700 | 32 GB | Q8_0 | 15 |
| Arc Pro B70 | 32 GB | Q8_0 | 14 |
| M5 Ultra | 512 GB | FP16 | 14 |
| Arc Pro B60 | 24 GB | Q6_K | 14 |
| M5 Max (18-core CPU / 32-core GPU) | 36 GB | Q6_K | 13 |
| M4 Max (14-core CPU / 32-core GPU) | 36 GB | Q6_K | 12 |
| M3 Ultra | 512 GB | FP16 | 9.7 |
| M5 Max (18-core CPU / 40-core GPU) | 128 GB | FP16 | 7.3 |
| M5 Pro | 64 GB | Q8_0 | 6.9 |
| M4 Max (16-core CPU / 40-core GPU) | 128 GB | FP16 | 6.5 |
| M4 Pro | 64 GB | Q8_0 | 6.1 |
| M6 (24GB/32GB configurations, 170 GB/s) | 32 GB | Q6_K | 4.9 |
| M5 | 32 GB | Q6_K | 4.4 |
| RTX Spark (N1X) laptops & compact desktops | 128 GB | FP16 | 3.3 |
| DGX Spark | 128 GB | FP16 | 3.0 |
| Ryzen AI Max+ PRO 495 (Gorgon Halo) systems | 192 GB | FP16 | 3.0 |
| Ryzen AI Max+ 395 (Strix Halo) systems | 128 GB | FP16 | 2.8 |
| GeForce RTX 3060 (12 GB) | 12 GB | Doesn't fit | – |
| GeForce RTX 4060 Ti 16GB | 16 GB | Doesn't fit | – |
| GeForce RTX 4080 SUPER | 16 GB | Doesn't fit | – |
| GeForce RTX 5060 Ti 16GB | 16 GB | Doesn't fit | – |
| GeForce RTX 5070 | 12 GB | Doesn't fit | – |
| GeForce RTX 5070 Ti | 16 GB | Doesn't fit | – |
| GeForce RTX 5080 | 16 GB | Doesn't fit | – |
| Radeon RX 9060 XT 16GB | 16 GB | Doesn't fit | – |
| Radeon RX 9070 XT | 16 GB | Doesn't fit | – |
| Arc B580 | 12 GB | Doesn't fit | – |
| M6 (16GB configuration, 153 GB/s) | 16 GB | Doesn't fit | – |

## Frequently asked questions

### How much VRAM does Gemma 3 27B IT need?

With an 8K context, about 17.5 GiB at Q4_K_M and 29.6 GiB at Q8_0 (weights + KV cache + runtime overhead). Longer contexts need more memory for the KV cache.

### Can Gemma 3 27B IT run on an GeForce RTX 4090?

Yes. Gemma 3 27B IT fits on the GeForce RTX 4090 (24 GB) at Q6_K with an 8K context, using about 23.3 GiB, at an estimated 30 tokens/s.

### Can I run Gemma 3 27B IT on a Mac?

Yes. Gemma 3 27B IT fits on the M3 Ultra (96 GB) at FP16 with an 8K context, using about 54.3 GiB, at an estimated 9.7 tokens/s.

### Can Gemma 3 27B IT be used commercially?

Gemma 3 27B IT is released under Gemma Terms of Use. The licence allows commercial use (read its conditions).

## Sources

- https://ai.google.dev/gemma/docs/releases
- https://ai.google.dev/gemma/terms
- https://ai.google.dev/gemma/docs/core/model_card_3
- https://huggingface.co/unsloth/gemma-3-27b-it/resolve/main/config.json
- https://huggingface.co/google/gemma-3-27b-it
- https://ollama.com/library/gemma3:27b/blobs/3eb75691d2db
- https://huggingface.co/api/models/google/gemma-3-27b-it?expand[]=safetensors
- https://ollama.com/library/gemma3/tags