Gemma · 9B to 32B
Gemma 3 27B Instruct
27.43 billion parameters across 62 layers. At four-bit precision the weights alone come to 15.42 GB, before any conversation is loaded. 51 of its 62 layers use a sliding window of 1k tokens, so only 11 keep growing with the conversation.
About Gemma 3 27B Instruct
The largest Gemma 3, with image input, a 128k window and the Gemma license. Most of its layers use sliding-window attention, so its cache stays small at long context.
Gemma 3 27B Instruct is a dense model with 27.43B parameters, built from 62 layers with a hidden size of 5376. Its repository was published on 2025-03-01. The license listed on Hugging Face is gemma. 51 of its 62 layers use a sliding window of 1k tokens, so only 11 keep growing with the conversation.
Memory needed, by quantization
At an 8k context window. Quality is our rough ranking of how much the compression costs you: anything at or above Q5 is hard to tell apart from the original in normal use.
| Quantization | File size | KV cache | Total VRAM | Quality | What it costs you |
|---|---|---|---|---|---|
| Q8_0 | 27.14 GB | 1.09 GB | 28.68 GB | 99% | Lossless in practice. Use it when the memory is there. |
| Q6_K | 20.95 GB | 1.09 GB | 22.48 GB | 98% | Very close to Q8 for two thirds of the size. |
| Q5_K_M | 18.17 GB | 1.09 GB | 19.71 GB | 96% | The quality-first choice when Q6 will not fit. |
| Q4_K_M | 15.42 GB | 1.09 GB | 16.96 GB | 93% | The default. Best size-to-quality ratio for local use. |
| IQ4_XS | 13.57 GB | 1.09 GB | 15.11 GB | 91% | Importance-matrix 4-bit. Q4_K_S quality, smaller file. |
| Q3_K_M | 12.49 GB | 1.09 GB | 14.02 GB | 86% | Degradation starts to show. A way to fit one size up. |
| IQ3_XXS | 9.77 GB | 1.09 GB | 11.31 GB | 79% | Aggressive. Only worth it on very large models. |
What a longer conversation costs
Same model at Q4_K_M, only the context window changes. 51 of its 62 layers use sliding-window attention, so their cache stops growing past 1k tokens and only the other 11 keep growing.
| Context | KV cache | Total VRAM | Fits in 8 GB | Fits in 12 GB | Fits in 24 GB |
|---|---|---|---|---|---|
| 4k | 0.74 GB | 16.62 GB | no | no | yes |
| 8k | 1.09 GB | 16.96 GB | no | no | yes |
| 16k | 1.77 GB | 17.65 GB | no | no | yes |
| 32k | 3.15 GB | 19.02 GB | no | no | yes |
| 64k | 5.9 GB | 21.77 GB | no | no | yes |
| 128k | 11.4 GB | 27.27 GB | no | no | no |
The "fits" columns allow for the roughly 0.8 GB Windows keeps for the desktop.
Which hardware runs Gemma 3 27B Instruct
60 of 118 consumer devices run it at a quantization worth using, at an 8k context window. Another 16 can load it only by compressing the weights far enough to damage the model, marked with a warning below.
| Device | Memory | Verdict | Quantization | Used | Speed |
|---|---|---|---|---|---|
| GeForce RTX 5090 NVIDIA | 32 GB | Runs great | Q8_0 | 28.68 GB | 46.7 t/s |
| RTX PRO 6000 Blackwell NVIDIA | 96 GB | Runs great | Q8_0 | 28.68 GB | 46.7 t/s |
| RTX 6000 Ada Generation NVIDIA | 48 GB | Runs great | Q8_0 | 28.68 GB | 25 t/s |
| RTX A6000 NVIDIA | 48 GB | Runs well | Q8_0 | 28.68 GB | 20 t/s |
| Apple M3 Ultra 96GB Apple | 96 GB | Runs well | Q8_0 | 28.65 GB | 19.7 t/s |
| Apple M3 Ultra 256GB Apple | 256 GB | Runs well | Q8_0 | 28.65 GB | 19.7 t/s |
| Apple M3 Ultra 512GB Apple | 512 GB | Runs well | Q8_0 | 28.65 GB | 19.7 t/s |
| Apple M1 Ultra 64GB Apple | 64 GB | Runs well | Q8_0 | 28.65 GB | 19.3 t/s |
| Apple M1 Ultra 128GB Apple | 128 GB | Runs well | Q8_0 | 28.65 GB | 19.3 t/s |
| Apple M2 Ultra 64GB Apple | 64 GB | Runs well | Q8_0 | 28.65 GB | 19.3 t/s |
| Apple M2 Ultra 128GB Apple | 128 GB | Runs well | Q8_0 | 28.65 GB | 19.3 t/s |
| Apple M2 Ultra 192GB Apple | 192 GB | Runs well | Q8_0 | 28.65 GB | 19.3 t/s |
| Radeon PRO W7900 AMD | 48 GB | Runs well | Q8_0 | 28.68 GB | 19.1 t/s |
| GeForce RTX 4090 NVIDIA | 24 GB | Just fits | Q6_K | 22.48 GB | 33.7 t/s |
| GeForce RTX 3090 Ti NVIDIA | 24 GB | Just fits | Q6_K | 22.48 GB | 33.7 t/s |
| GeForce RTX 3090 NVIDIA | 24 GB | Just fits | Q6_K | 22.48 GB | 31.3 t/s |
| GeForce RTX 5090 Laptop NVIDIA | 24 GB | Just fits | Q6_K | 22.48 GB | 29.9 t/s |
| Radeon RX 7900 XTX AMD | 24 GB | Just fits | Q6_K | 22.48 GB | 27.2 t/s |
| RTX A5000 NVIDIA | 24 GB | Just fits | Q6_K | 22.48 GB | 25.6 t/s |
| Radeon RX 7900 XT AMD | 20 GB | Runs great | Q4_K_M | 16.96 GB | 30.2 t/s |
| GeForce RTX 5080 NVIDIA | 16 GB | Just fits | IQ4_XS | 15.11 GB | 48.2 t/s |
| GeForce RTX 5070 Ti NVIDIA | 16 GB | Just fits | IQ4_XS | 15.11 GB | 45 t/s |
| GeForce RTX 5080 Laptop NVIDIA | 16 GB | Just fits | IQ4_XS | 15.11 GB | 38.6 t/s |
| GeForce RTX 4080 SUPER NVIDIA | 16 GB | Just fits | IQ4_XS | 15.11 GB | 36.9 t/s |
| GeForce RTX 4080 NVIDIA | 16 GB | Just fits | IQ4_XS | 15.11 GB | 36 t/s |
Show 35 more devices
| GeForce RTX 4070 Ti SUPER NVIDIA | 16 GB | Just fits | IQ4_XS | 15.11 GB | 33.7 t/s |
| GeForce RTX 4090 Laptop NVIDIA | 16 GB | Just fits | IQ4_XS | 15.11 GB | 28.9 t/s |
| Radeon RX 9070 XT AMD | 16 GB | Just fits | IQ4_XS | 15.11 GB | 27.5 t/s |
| Radeon RX 9070 AMD | 16 GB | Just fits | IQ4_XS | 15.11 GB | 27.5 t/s |
| Radeon RX 7800 XT AMD | 16 GB | Just fits | IQ4_XS | 15.11 GB | 26.6 t/s |
| Radeon RX 7900 GRE AMD | 16 GB | Just fits | IQ4_XS | 15.11 GB | 24.5 t/s |
| GeForce RTX 5060 Ti 16GB NVIDIA | 16 GB | Just fits | IQ4_XS | 15.11 GB | 22.5 t/s |
| RTX A4000 NVIDIA | 16 GB | Just fits | IQ4_XS | 15.11 GB | 22.5 t/s |
| Radeon RX 6900 XT AMD | 16 GB | Just fits | IQ4_XS | 15.11 GB | 21.8 t/s |
| Radeon RX 6800 AMD | 16 GB | Just fits | IQ4_XS | 15.11 GB | 21.8 t/s |
| Arc A770 16GB Intel | 16 GB | Just fits | IQ4_XS | 15.11 GB | 21.7 t/s |
| Apple M4 Max 48GB Apple | 48 GB | Runs well | Q8_0 | 28.65 GB | 13.1 t/s |
| Apple M4 Max 64GB Apple | 64 GB | Runs well | Q8_0 | 28.65 GB | 13.1 t/s |
| Apple M4 Max 128GB Apple | 128 GB | Runs well | Q8_0 | 28.65 GB | 13.1 t/s |
| Apple M1 Max 64GB Apple | 64 GB | Runs well | Q6_K | 22.45 GB | 12.3 t/s |
| Apple M2 Max 64GB Apple | 64 GB | Runs well | Q6_K | 22.45 GB | 12.3 t/s |
| Apple M2 Max 96GB Apple | 96 GB | Runs well | Q6_K | 22.45 GB | 12.3 t/s |
| Apple M3 Max 36GB Apple | 36 GB | Runs well | Q6_K | 22.45 GB | 12.3 t/s |
| Apple M3 Max 48GB Apple | 48 GB | Runs well | Q6_K | 22.45 GB | 12.3 t/s |
| Apple M3 Max 64GB Apple | 64 GB | Runs well | Q6_K | 22.45 GB | 12.3 t/s |
| Apple M3 Max 96GB Apple | 96 GB | Runs well | Q6_K | 22.45 GB | 12.3 t/s |
| Apple M3 Max 128GB Apple | 128 GB | Runs well | Q6_K | 22.45 GB | 12.3 t/s |
| GeForce RTX 4060 Ti 16GB NVIDIA | 16 GB | Just fits | IQ4_XS | 15.11 GB | 14.5 t/s |
| NVIDIA DGX Spark 128GB NVIDIA | 128 GB | Runs well | Q5_K_M | 19.71 GB | 10.4 t/s |
| Apple M4 Pro 48GB Apple | 48 GB | Runs well | Q4_K_M | 16.93 GB | 11.2 t/s |
| Apple M4 Pro 64GB Apple | 64 GB | Runs well | Q4_K_M | 16.93 GB | 11.2 t/s |
| Radeon RX 7600 XT AMD | 16 GB | Just fits | IQ4_XS | 15.11 GB | 12.3 t/s |
| Jetson AGX Orin 32GB NVIDIA | 32 GB | Runs well | IQ4_XS | 15.11 GB | 10.3 t/s |
| Jetson AGX Orin 64GB NVIDIA | 64 GB | Runs well | IQ4_XS | 15.11 GB | 10.3 t/s |
| Ryzen AI Max+ 395 32GB AMD | 32 GB | Runs well | Q3_K_M | 14.02 GB | 10.7 t/s |
| Ryzen AI Max+ 395 64GB AMD | 64 GB | Runs well | Q3_K_M | 14.02 GB | 10.7 t/s |
| Ryzen AI Max+ 395 96GB AMD | 96 GB | Runs well | Q3_K_M | 14.02 GB | 10.7 t/s |
| Ryzen AI Max+ 395 128GB AMD | 128 GB | Runs well | Q3_K_M | 14.02 GB | 10.7 t/s |
| Apple M1 Pro 32GB Apple | 32 GB | Runs well | Q3_K_M | 13.99 GB | 10 t/s |
| Apple M2 Pro 32GB Apple | 32 GB | Runs well | Q3_K_M | 13.99 GB | 10 t/s |
How it compares
Other models in the same family or of a similar size, on a 12 GB card and a 24 GB card, each at the best quantization that fits.
| Model | Weights at 4 bits | GeForce RTX 3060 12GB | GeForce RTX 4090 |
|---|---|---|---|
| Gemma 3 27B Instruct | 15.4 GB | does not fit well | 33.7 t/s at Q6_K |
| Gemma 4 26B A4B Instruct | 14.5 GB | 37.6 t/s at Q4_K_M | 117 t/s at Q6_K |
| Gemma 4 31B Instruct | 17.6 GB | does not fit well | 35.5 t/s at Q4_K_M |
| Gemma 3 12B Instruct | 6.9 GB | 26.2 t/s at Q6_K | 57.6 t/s at Q8_0 |
| GPT OSS 20B | 11.3 GB | 65.7 t/s at MXFP4 | 180.1 t/s at MXFP4 |
| Qwen3 32B | 18.4 GB | does not fit well | 37.1 t/s at Q4_K_M |
What to buy to run Gemma 3 27B Instruct
The cheapest hardware that runs it at a quantization worth using and a speed you would not resent, at an 8k context window.
Cheapest that works
Arc A770 16GB
$280 used
IQ4_XS · uses 15.11 GB of its 16 GB · about 21.7 tokens/s
Best value
Radeon RX 6800
$330 used
IQ4_XS · about 21.8 tokens/s · 512 GB/s
| Device | Price | Memory | Runs it at | Speed | |
|---|---|---|---|---|---|
| Arc A770 16GB Intel | $280 used | 16 GB | IQ4_XS | 21.7 t/s | Check price |
| Radeon RX 6800 AMD | $330 used | 16 GB | IQ4_XS | 21.8 t/s | Check price |
| GeForce RTX 5060 Ti 16GB NVIDIA | $430 | 16 GB | IQ4_XS | 22.5 t/s | Check price |
| Radeon RX 7800 XT AMD | $480 | 16 GB | IQ4_XS | 26.6 t/s | Check price |
| Radeon RX 9070 AMD | $560 | 16 GB | IQ4_XS | 27.5 t/s | Check price |
| Radeon RX 7900 XT AMD | $650 | 20 GB | Q4_K_M | 30.2 t/s | Check price |
| GeForce RTX 3090 NVIDIA | $700 used | 24 GB | Q6_K | 31.3 t/s | Check price |
| GeForce RTX 4070 Ti SUPER NVIDIA | $750 | 16 GB | IQ4_XS | 33.7 t/s | Check price |
Indicative prices reviewed 2026-09-01; used prices are marketplace typical. Options that cost more than a cheaper one with no more memory and no more speed are hidden. Price links are Amazon affiliate links. Change the standard, the context window or the budget in the buying tool.
Where it sits in the catalog
Source: google/gemma-3-27b-it on Hugging Face. Downloaded 441k times in the last month. Published 2025-03-01. Architecture figures are read from the repository's own configuration file, so they move when the model does.
Gemma 4 26B A4B Instruct
25.81B (3.74B active) · 256k context
Gemma 4 31B Instruct
31.27B · 256k context
Gemma 3 12B Instruct
12.19B · 128k context
Gemma 4 12B Instruct
11.96B · 256k context
Gemma 2 9B Instruct
9.24B · 8k context
Gemma 4 E4B Instruct
8B · 128k context
Gemma 4 E2B Instruct
5.12B · 128k context
Gemma 3 4B Instruct
4.3B · 128k context
Questions people ask about Gemma 3 27B Instruct
How much VRAM does Gemma 3 27B Instruct need?
At Q4_K_M with an 8k context window, Gemma 3 27B Instruct needs about 16.96 GB: 15.42 GB of weights, 1.09 GB of KV cache and 0.45 GB of overhead. At the highest-quality file it needs about 28.68 GB.
What is the cheapest GPU that runs Gemma 3 27B Instruct?
The smallest consumer device in our catalog that fits it is the GeForce RTX 5080 with 16 GB, running it at IQ4_XS at roughly 48.2 tokens per second.
How much does a longer context window add for Gemma 3 27B Instruct?
On the memory side, a 32k window adds 3.1 GB of KV cache on top of the weights, and the cache grows more slowly than a conventional model because of its attention design. The longest window the model supports natively is 128k tokens.