Gemma · 9B to 32B
Gemma 4 31B Instruct
31.27 billion parameters across 60 layers. At four-bit precision the weights alone come to 17.58 GB, before any conversation is loaded. This one is cache-hungry: a 32k window adds 10.8 GB on top, so context length matters more here than the model size suggests.
About Gemma 4 31B Instruct
Gemma 4 31B Instruct is a dense model with 31.27B parameters, built from 60 layers with a hidden size of 5376. Its repository was published on 2026-03-11. The license listed on Hugging Face is apache-2.0. 50 of its 60 layers use a sliding window of 1k tokens, so only 10 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 | 30.94 GB | 3.28 GB | 34.67 GB | 99% | Lossless in practice. Use it when the memory is there. |
| Q6_K | 23.88 GB | 3.28 GB | 27.61 GB | 98% | Very close to Q8 for two thirds of the size. |
| Q5_K_M | 20.71 GB | 3.28 GB | 24.44 GB | 96% | The quality-first choice when Q6 will not fit. |
| Q4_K_M | 17.58 GB | 3.28 GB | 21.31 GB | 93% | The default. Best size-to-quality ratio for local use. |
| IQ4_XS | 15.47 GB | 3.28 GB | 19.2 GB | 91% | Importance-matrix 4-bit. Q4_K_S quality, smaller file. |
| Q3_K_M | 14.23 GB | 3.28 GB | 17.96 GB | 86% | Degradation starts to show. A way to fit one size up. |
| IQ3_XXS | 11.14 GB | 3.28 GB | 14.87 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. 50 of its 60 layers use sliding-window attention, so their cache stops growing past 1k tokens and only the other 10 keep growing.
| Context | KV cache | Total VRAM | Fits in 8 GB | Fits in 12 GB | Fits in 24 GB |
|---|---|---|---|---|---|
| 4k | 2.03 GB | 20.06 GB | no | no | yes |
| 8k | 3.28 GB | 21.31 GB | no | no | yes |
| 16k | 5.78 GB | 23.81 GB | no | no | no |
| 32k | 10.78 GB | 28.81 GB | no | no | no |
| 64k | 20.78 GB | 38.81 GB | no | no | no |
| 128k | 40.78 GB | 58.81 GB | no | no | no |
The "fits" columns allow for the roughly 0.8 GB Windows keeps for the desktop.
Which hardware runs Gemma 4 31B Instruct
37 of 118 consumer devices run it at a quantization worth using, at an 8k context window. Another 18 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 |
|---|---|---|---|---|---|
| RTX PRO 6000 Blackwell NVIDIA | 96 GB | Runs great | Q8_0 | 34.67 GB | 38.5 t/s |
| RTX 6000 Ada Generation NVIDIA | 48 GB | Runs well | Q8_0 | 34.67 GB | 20.6 t/s |
| GeForce RTX 5090 NVIDIA | 32 GB | Runs great | Q6_K | 27.61 GB | 48.5 t/s |
| RTX A6000 NVIDIA | 48 GB | Runs well | Q8_0 | 34.67 GB | 16.5 t/s |
| Apple M3 Ultra 96GB Apple | 96 GB | Runs well | Q8_0 | 34.64 GB | 16.3 t/s |
| Apple M3 Ultra 256GB Apple | 256 GB | Runs well | Q8_0 | 34.64 GB | 16.3 t/s |
| Apple M3 Ultra 512GB Apple | 512 GB | Runs well | Q8_0 | 34.64 GB | 16.3 t/s |
| Apple M1 Ultra 64GB Apple | 64 GB | Runs well | Q8_0 | 34.64 GB | 15.9 t/s |
| Apple M1 Ultra 128GB Apple | 128 GB | Runs well | Q8_0 | 34.64 GB | 15.9 t/s |
| Apple M2 Ultra 64GB Apple | 64 GB | Runs well | Q8_0 | 34.64 GB | 15.9 t/s |
| Apple M2 Ultra 128GB Apple | 128 GB | Runs well | Q8_0 | 34.64 GB | 15.9 t/s |
| Apple M2 Ultra 192GB Apple | 192 GB | Runs well | Q8_0 | 34.64 GB | 15.9 t/s |
| Radeon PRO W7900 AMD | 48 GB | Runs well | Q8_0 | 34.67 GB | 15.8 t/s |
| GeForce RTX 4090 NVIDIA | 24 GB | Runs great | Q4_K_M | 21.31 GB | 35.5 t/s |
| GeForce RTX 3090 Ti NVIDIA | 24 GB | Runs great | Q4_K_M | 21.31 GB | 35.5 t/s |
| GeForce RTX 3090 NVIDIA | 24 GB | Runs great | Q4_K_M | 21.31 GB | 33 t/s |
| GeForce RTX 5090 Laptop NVIDIA | 24 GB | Runs great | Q4_K_M | 21.31 GB | 31.6 t/s |
| Radeon RX 7900 XTX AMD | 24 GB | Runs great | Q4_K_M | 21.31 GB | 28.7 t/s |
| RTX A5000 NVIDIA | 24 GB | Runs great | Q4_K_M | 21.31 GB | 27.1 t/s |
| Apple M4 Max 48GB Apple | 48 GB | Runs well | Q8_0 | 34.64 GB | 10.8 t/s |
| Apple M4 Max 64GB Apple | 64 GB | Runs well | Q8_0 | 34.64 GB | 10.8 t/s |
| Apple M4 Max 128GB Apple | 128 GB | Runs well | Q8_0 | 34.64 GB | 10.8 t/s |
| Radeon RX 7900 XT AMD | 20 GB | Runs great | Q3_K_M | 17.96 GB | 28.5 t/s |
| Apple M1 Max 64GB Apple | 64 GB | Runs well | Q6_K | 27.58 GB | 10 t/s |
| Apple M2 Max 64GB Apple | 64 GB | Runs well | Q6_K | 27.58 GB | 10 t/s |
Show 35 more devices
| Apple M2 Max 96GB Apple | 96 GB | Runs well | Q6_K | 27.58 GB | 10 t/s |
| Apple M3 Max 48GB Apple | 48 GB | Runs well | Q6_K | 27.58 GB | 10 t/s |
| Apple M3 Max 64GB Apple | 64 GB | Runs well | Q6_K | 27.58 GB | 10 t/s |
| Apple M3 Max 96GB Apple | 96 GB | Runs well | Q6_K | 27.58 GB | 10 t/s |
| Apple M3 Max 128GB Apple | 128 GB | Runs well | Q6_K | 27.58 GB | 10 t/s |
| NVIDIA DGX Spark 128GB NVIDIA | 128 GB | Runs well | IQ4_XS | 19.2 GB | 10.7 t/s |
| GeForce RTX 5080 NVIDIA | 16 GB | Just fits | IQ3_XXS ! | 14.87 GB | 49 t/s |
| GeForce RTX 5070 Ti NVIDIA | 16 GB | Just fits | IQ3_XXS ! | 14.87 GB | 45.7 t/s |
| GeForce RTX 5080 Laptop NVIDIA | 16 GB | Just fits | IQ3_XXS ! | 14.87 GB | 39.2 t/s |
| GeForce RTX 4080 SUPER NVIDIA | 16 GB | Just fits | IQ3_XXS ! | 14.87 GB | 37.6 t/s |
| GeForce RTX 4080 NVIDIA | 16 GB | Just fits | IQ3_XXS ! | 14.87 GB | 36.6 t/s |
| GeForce RTX 4070 Ti SUPER NVIDIA | 16 GB | Just fits | IQ3_XXS ! | 14.87 GB | 34.3 t/s |
| GeForce RTX 4090 Laptop NVIDIA | 16 GB | Just fits | IQ3_XXS ! | 14.87 GB | 29.4 t/s |
| Radeon RX 9070 XT AMD | 16 GB | Just fits | IQ3_XXS ! | 14.87 GB | 27.9 t/s |
| Radeon RX 9070 AMD | 16 GB | Just fits | IQ3_XXS ! | 14.87 GB | 27.9 t/s |
| Radeon RX 7800 XT AMD | 16 GB | Just fits | IQ3_XXS ! | 14.87 GB | 27 t/s |
| Radeon RX 7900 GRE AMD | 16 GB | Just fits | IQ3_XXS ! | 14.87 GB | 24.9 t/s |
| GeForce RTX 5060 Ti 16GB NVIDIA | 16 GB | Just fits | IQ3_XXS ! | 14.87 GB | 22.9 t/s |
| RTX A4000 NVIDIA | 16 GB | Just fits | IQ3_XXS ! | 14.87 GB | 22.9 t/s |
| Radeon RX 6900 XT AMD | 16 GB | Just fits | IQ3_XXS ! | 14.87 GB | 22.2 t/s |
| Radeon RX 6800 AMD | 16 GB | Just fits | IQ3_XXS ! | 14.87 GB | 22.2 t/s |
| Arc A770 16GB Intel | 16 GB | Just fits | IQ3_XXS ! | 14.87 GB | 22.1 t/s |
| Apple M4 Pro 48GB Apple | 48 GB | Runs well | Q3_K_M | 17.93 GB | 10.6 t/s |
| Apple M4 Pro 64GB Apple | 64 GB | Runs well | Q3_K_M | 17.93 GB | 10.6 t/s |
| GeForce RTX 4060 Ti 16GB NVIDIA | 16 GB | Just fits | IQ3_XXS ! | 14.87 GB | 14.7 t/s |
| Jetson AGX Orin 64GB NVIDIA | 64 GB | Fits, but slow | Q8_0 | 34.67 GB | 4.4 t/s |
| Ryzen AI Max+ 395 64GB AMD | 64 GB | Fits, but slow | Q8_0 | 34.67 GB | 4.2 t/s |
| Ryzen AI Max+ 395 96GB AMD | 96 GB | Fits, but slow | Q8_0 | 34.67 GB | 4.2 t/s |
| Ryzen AI Max+ 395 128GB AMD | 128 GB | Fits, but slow | Q8_0 | 34.67 GB | 4.2 t/s |
| Radeon RX 7600 XT AMD | 16 GB | Just fits | IQ3_XXS ! | 14.87 GB | 12.5 t/s |
| Apple M4 Max 36GB Apple | 36 GB | Just fits | n/a | 27.58 GB | 13.7 t/s |
| Apple M1 Pro 16GB Apple | 16 GB | Just fits | n/a ! | 11.2 GB | 12.6 t/s |
| Apple M2 Pro 16GB Apple | 16 GB | Just fits | n/a ! | 11.2 GB | 12.6 t/s |
| Apple M4 Pro 24GB Apple | 24 GB | Just fits | n/a | 17.93 GB | 10.6 t/s |
| Apple M1 Max 32GB Apple | 32 GB | Just fits | n/a | 27.58 GB | 10 t/s |
Can my GPU run Gemma 4 31B Instruct?
A direct answer for popular devices, each with the settings we would use.
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 4 31B Instruct | 17.6 GB | does not fit well | 35.5 t/s at Q4_K_M |
| 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 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 4 31B 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
GeForce RTX 3090
$700 used
Q4_K_M · uses 21.31 GB of its 24 GB · about 33 tokens/s
Best value
GeForce RTX 3090 Ti
$800 used
Q4_K_M · about 35.5 tokens/s · 1008 GB/s
If it just has to run
Radeon RX 7900 XT
$650
Q3_K_M · about 28.5 tokens/s. Cheaper because it runs the model at a lower precision.
| Device | Price | Memory | Runs it at | Speed | |
|---|---|---|---|---|---|
| GeForce RTX 3090 NVIDIA | $700 used | 24 GB | Q4_K_M | 33 t/s | Check price |
| GeForce RTX 3090 Ti NVIDIA | $800 used | 24 GB | Q4_K_M | 35.5 t/s | Check price |
| GeForce RTX 5090 NVIDIA | $2,200 | 32 GB | Q6_K | 48.5 t/s | Check price |
| Apple M1 Ultra 64GB Apple · whole machine | $2,400 | 64 GB | Q8_0 | 15.9 t/s | Check price |
| Apple M1 Ultra 128GB Apple · whole machine | $3,400 | 128 GB | Q8_0 | 15.9 t/s | Check price |
| Apple M3 Ultra 256GB Apple · whole machine | $5,599 | 256 GB | Q8_0 | 16.3 t/s | Check price |
| Apple M3 Ultra 512GB Apple · whole machine | $9,499 | 512 GB | Q8_0 | 16.3 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-4-31B-it on Hugging Face. Downloaded 8.3M times in the last month. Published 2026-03-11. Architecture figures are read from the repository's own configuration file, so they move when the model does.
Gemma 3 27B Instruct
27.43B · 128k context
Gemma 4 26B A4B Instruct
25.81B (3.74B active) · 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 4 31B Instruct
How much VRAM does Gemma 4 31B Instruct need?
At Q4_K_M with an 8k context window, Gemma 4 31B Instruct needs about 21.31 GB: 17.58 GB of weights, 3.28 GB of KV cache and 0.45 GB of overhead. At the highest-quality file it needs about 34.67 GB.
What is the cheapest GPU that runs Gemma 4 31B Instruct?
The smallest consumer device in our catalog that fits it is the Radeon RX 7900 XT with 20 GB, running it at Q3_K_M at roughly 28.5 tokens per second.
How much does a longer context window add for Gemma 4 31B Instruct?
On the memory side, a 32k window adds 10.8 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 256k tokens.