GoogleGemma 4Gemma4ForConditionalGeneration

Gemma 4 31B IT — Hardware Requirements & GPU Compatibility

VisionFunctions

Gemma 4 31B IT is Google's 31-billion-parameter instruction-tuned model in the Gemma 4 lineup, built to handle both text and image input in a single pass. As a vision-capable model, it can describe, compare, or reason about images alongside written prompts, making it suitable for multimodal chat and document-understanding tasks. At this parameter count, local inference calls for quantization and a fairly capable GPU; it fits on a single high-end consumer or workstation card rather than lower-end hardware. The model supports a 256K token context window, enough for long documents, transcripts, or multi-turn conversations without aggressive truncation. It is released under the Apache 2.0 license, allowing unrestricted commercial and research use, and reflects Google's continued push toward mid-sized, multimodal open-weight models following the earlier Gemma generations.

31.3B params · dense262K contextApache 2.0Mar 20267.5M downloads4.3K likes2.9M quant downloads

Based on Gemma 4 31B

Your hardware

Verdicts below are for the reference RTX 4090. Set yours to personalise the page.

Detection reads your browser's GPU name with no data sent anywhere.

Decent on a 24 GB GPU

Runs from 12 GB (IQ2_XXS · 16K)Best on 24 GB (Q4_K_M · ~32 tok/s · 32K)Macs from 16 GB (IQ2_XS · ~13 tok/s)

Score 106.7#42 of 106 open#4 of 36 that fit 24 GBApache 2.0

Reference GPU · RTX 4090 24 GB

Estimates, ±20% · how we estimate

llmrun Composite Score →

106.7

90% range 103.7–109.7
#42 of 106 open models
#4 of 36 that fit 24 GB

Benchmarks · tier chart ↓

Among 106 open modelsHow the Composite Score is built

Coding107.190% range 103–112
AgentsNot enough data yet
MathNot enough data yet
Science105.290% range 97–114
Reasoning111.190% range 104–119

Run Gemma 4 31B IT

One recommended way. Every other method, OS and quant is in the step-by-step guide.Each way opens a step-by-step guide for your OS and quant.

Ways to runOther ways

each opens step-by-step

Run in cloud

Your own private copy · Q4_K_M · 24 GB GPU

The easiest way to try this model
Running in a few minutes, easy setup
Chat right here or connect any app

$5+ bonus for new users

Run Gemma 4 31B IT

21.7 GB VRAM at 16K context · fits the RTX 4090 (2.3 GB free)

Works on macOS, Windows and Linux.

Install Ollama (Linux)
$curl -fsSL https://ollama.com/install.sh | sh
Install Ollama (macOS / Windows)
Download and run the app; it installs the ollama command.
Download Ollama ↗
Run Q4_K_M
$ollama run hf.co/unsloth/gemma-4-31B-it-GGUF:Q4_K_M
Raise the context window
Type it inside the chat. For the API server instead, start it with OLLAMA_CONTEXT_LENGTH=16384 ollama serve.
$/set parameter num_ctx 16384
A larger context needs more memory than the weights alone.
Use the API
Ollama serves its API on http://localhost:11434.

Works on macOS, Windows and Linux.

Install LM Studio
Desktop app for macOS (Apple silicon), Windows and Linux.
lmstudio.ai ↗
Download the model
In the app press Ctrl/Cmd + Shift + M and search google/gemma-4-31b, then pick Q4_K_M. Or use the CLI (run LM Studio once first):
$lms get google/gemma-4-31b@q4_k_m
Open on Hugging Face ↗
Load and chat
Open the model loader, select the model and set Context Length to 16384. Leave GPU offload at maximum if it fits.
Start the local server (optional)
$lms server start

Works on macOS, Windows and Linux.

Install llama.cpp (macOS)
MacPorts, Nix and conda-forge also work.
$brew install llama.cpp
Install options ↗
Install llama.cpp (Windows)
$winget install llama.cpp
Install llama.cpp (Linux)
Download the prebuilt llama-*-bin-ubuntu-* archive for your GPU backend (CPU, Vulkan) from the releases page, unpack it and run the binaries from that folder. Homebrew (brew install llama.cpp) and conda-forge also work on Linux.
llama.cpp releases ↗
Start the server
Downloads the model on first run. The chat UI and an OpenAI-compatible API are at http://localhost:8080.
$llama-server -hf unsloth/gemma-4-31B-it-GGUF:Q4_K_M -c 16384 -ngl 99 --port 8080

Works on macOS, Windows and Linux.

Option A: Docker Model Runner
Docker Desktop: Settings > AI > Enable Docker Model Runner. Docker Engine on Linux: install the docker-model-plugin package.
$docker model run hf.co/unsloth/gemma-4-31B-it-GGUF:Q4_K_M
Model Runner docs ↗
The quant tag after the repo is how Hugging Face cards write it; if your Docker version rejects it, drop the tag and pick the quant by repo.
Option B: llama.cpp server image (NVIDIA)
Needs the NVIDIA Container Toolkit. Models are cached in the llama-cache volume. Open http://localhost:8080.
$docker run --gpus all -p 8080:8080 -v llama-cache:/root/.cache ghcr.io/ggml-org/llama.cpp:server-cuda -hf unsloth/gemma-4-31B-it-GGUF:Q4_K_M -c 16384 -ngl 99 --host 0.0.0.0 --port 8080
Image tags (cuda13, vulkan, rocm) ↗
Cache path inside the image is not in the docs; remove the -v flag if the volume complains.

Linux only.

Install vLLM
Linux only; on Windows run it inside WSL2. NVIDIA GPU shown; AMD ROCm, Intel and CPU builds have their own install pages.
$uv venv --python 3.12 --seed
$source .venv/bin/activate
$uv pip install vllm --torch-backend=auto
vLLM installation ↗
Serve the model
OpenAI-compatible API at http://localhost:8000/v1.
$vllm serve google/gemma-4-31B-it --max-model-len 16384
Runs the original weights, not the Q4_K_M GGUF; needs ~69 GB of GPU memory.

macOS only.

MLX supports only macOS, so follow these steps for that setup.

Install mlx-lm
$pip install mlx-lm
Chat in the terminal
$mlx_lm.chat --model lmstudio-community/gemma-4-31B-it-MLX-4bit
Downloads the model on first run. The MLX repo has its own quantisation; it may not match the quant selected above.
Or start a local server
Listens on http://localhost:8080.
$mlx_lm.server --model lmstudio-community/gemma-4-31B-it-MLX-4bit

Works on macOS, Windows and Linux.

Install Jan
jan.ai ↗
Find the model
Open the Hub tab, paste this model ID in Search, and choose Q4_K_M if offered.
$unsloth/gemma-4-31B-it-GGUF
Use the local API (optional)
Settings > Local API Server > Start Server. It listens on http://127.0.0.1:1337.

Rent GPUs hourly to run Q4_K_M right there. Pricing comes from partner marketplaces and varies.

Available GPU rentals for Q4_K_M, prioritized by value
GPU	Speed	$/1M tokens	Rent
RTX 5090 (32 GB)CheapestCheapest10 GB free	~57 tok/s	$2.42	Deploy on Vast.aifrom $0.50/hr (opens in new tab)	32 GB · 10 GB free · ~57 tok/s · $2.42 / 1M tokens
RTX PRO 4500 (32 GB)10 GB free	~29 tok/s	$3.30	Deploy on RunPodfrom $0.34/hr (opens in new tab)	32 GB · 10 GB free · ~29 tok/s · $3.30 / 1M tokens
A40 (48 GB)26 GB free	~22 tok/s	$4.38	Deploy on RunPodfrom $0.35/hr (opens in new tab)	48 GB · 26 GB free · ~22 tok/s · $4.38 / 1M tokens
RTX PRO 4500 (32 GB)10 GB free	~29 tok/s	$3.45	Deploy on Vast.aifrom $0.36/hr (opens in new tab)	32 GB · 10 GB free · ~29 tok/s · $3.45 / 1M tokens
RTX A6000 (48 GB)26 GB free	~25 tok/s	$4.59	Deploy on Vast.aifrom $0.40/hr (opens in new tab)	48 GB · 26 GB free · ~25 tok/s · $4.59 / 1M tokens
A100 40GB18 GB free	~50 tok/s	$3.42	Deploy on Vast.aifrom $0.61/hr (opens in new tab)	18 GB free · ~50 tok/s · $3.42 / 1M tokens
L40 (48 GB)26 GB free	~28 tok/s	$6.97	Deploy on RunPodfrom $0.69/hr (opens in new tab)	48 GB · 26 GB free · ~28 tok/s · $6.97 / 1M tokens
RTX 5090 (32 GB)10 GB free	~57 tok/s	$3.36	Deploy on RunPodfrom $0.69/hr (opens in new tab)	32 GB · 10 GB free · ~57 tok/s · $3.36 / 1M tokens

Compare GPUs →

Your own private copy of Gemma 4 31B IT · Q4_K_M · OpenAI-compatible API

Open the llama.cpp listing on RunPod
Open on RunPodServerless (opens in new tab)
New to RunPod? Sign up and add credit first.
2
Fill in the form
Press Deploy and fill in:
Model
unsloth/gemma-4-31B-it-GGUF
Quantization
Q4_K_M
Context Size
16384
After pasting the model press Esc — don’t pick a suggestion. Press Next.
Create the endpoint
Keep the default 24 GB and 48 GB, then Create Endpoint. Right away: Manage → Edit endpoint → Max workers 1 → Save (changing settings later restarts the worker).
The first start downloads the model, usually 5–15 min. After that, replies after a pause come in seconds.
Create an API key
RunPod → Settings → API Keys → Create. Pick “Restricted” with access to this endpoint only.

Use it

Copy the endpoint URL from its page on RunPod (any of its links works) and paste it with the key. Both stay in this browser.

Endpoint URL

API key

Chats and keys stay only in this browser.

https://api.runpod.ai/v2/<ENDPOINT_ID>/openai/v1

Put your endpoint ID in place of <ENDPOINT_ID>. The API key is the one from step 2.

from openai import OpenAI
client = OpenAI(base_url="https://api.runpod.ai/v2/<ENDPOINT_ID>/openai/v1", api_key="rpa_...")
r = client.chat.completions.create(
    model="any",
    messages=[{"role": "user", "content": "Hello!"}])
print(r.choices[0].message.content)

Works with any OpenAI client.

https://api.runpod.ai/v2/<ENDPOINT_ID>/openai/v1

Put your endpoint ID in place of <ENDPOINT_ID>. The API key is the one from step 2.

Free on Windows, macOS, Linux, iOS, Android and the web.

Open the web app ↗ (opens in a new tab)Download ↗ (opens in a new tab)

1. Settings → Model Provider → Add → Add Custom Provider
2. API Mode OpenAI API Compatible; paste the API key and the base URL as API Host
3. Model → Fetch, pick the model
4. New chat → pick the provider → start chatting

Free and open source for Windows, macOS and Linux.

Download Jan ↗ (opens in a new tab)

1.Settings → next to Model Providers press +
2.API format OpenAI-compatible; Base URL is the base URL above (with /v1); API key is your key; Create
3.If no model appears, click + in Models and type any
4.New chat → pick the model → start chatting

How Much VRAM Does Gemma 4 31B IT Need?

Select a quantization level; the listed hardware adheres to it.

VRAM	+32K	+128K	File	Quality	On RTX 4090 (reference)
10.2 GB	13.7 GB	26.3 GB	8.5 GB	2-bit · i-quant extreme compression — Importance-weighted 2-bit, extreme compression — significant quality loss	Fits: 64K context · ~64 tok/s
11 GB	14.4 GB	27.1 GB	11.5 GB	2-bit · i-quant extra small — Importance-weighted 2-bit, extra small	Fits: 64K context · ~60 tok/s
11.4 GB	14.8 GB	27.5 GB	12.1 GB	2-bit · i-quant small — Importance-weighted 2-bit, small	Fits: 64K context · ~58 tok/s
12.2 GB	15.6 GB	28.3 GB	10.8 GB	2-bit · i-quant medium — Importance-weighted 2-bit, medium	Fits: 64K context · ~54 tok/s
13.7 GB	17.2 GB	29.9 GB	11.8 GB	3-bit · i-quant — Importance-weighted 3-bit	Fits: 64K context · ~48 tok/s
14.1 GB	17.6 GB	30.3 GB	11 GB	2-bit · small K-quant — 2-bit small K-quant	Fits: 64K context · ~46 tok/s
14.5 GB	18 GB	30.6 GB	13.8 GB	3-bit · i-quant extra small — Importance-weighted 3-bit, extra small	Fits: 64K context · ~45 tok/s
14.9 GB	18.4 GB	31 GB	13.8 GB	3-bit · i-quant small — Importance-weighted 3-bit, small	Fits: 64K context · ~44 tok/s
14.9 GB	18.4 GB	31 GB	12.6 GB	2-bit · K-quant — 2-bit quantization with K-quant improvements	Fits: 64K context · ~44 tok/s
15.3 GB	18.7 GB	31.4 GB	13.2 GB	3-bit small — 3-bit small quantization	Fits: 64K context · ~43 tok/s
15.7 GB	19.1 GB	31.8 GB	15.1 GB	3-bit · i-quant medium — Importance-weighted 3-bit, medium	Fits: 64K context · ~42 tok/s
16.9 GB	20.3 GB	33 GB	14.7 GB	3-bit medium — 3-bit medium quantization	Fits: 32K context · ~39 tok/s
17.3 GB	20.7 GB	33.4 GB	17.3 GB	4-bit · legacy — 4-bit legacy quantization	Fits: 32K context · ~38 tok/s
17.6 GB	21.1 GB	33.8 GB	16.8 GB	3-bit large — 3-bit large quantization	Fits: 32K context · ~37 tok/s
18.4 GB	21.9 GB	34.6 GB	16.4 GB	4-bit · i-quant compact — Importance-weighted 4-bit, compact	Fits: 32K context · ~36 tok/s
19.2 GB	22.7 GB	35.3 GB	19.1 GB	4-bit · legacy+ — 4-bit legacy quantization with offset	Fits: 32K context · ~34 tok/s
19.2 GB	22.7 GB	35.3 GB	17.4 GB	4-bit small — 4-bit small quantization	Fits: 32K context · ~34 tok/s
19.2 GB	22.7 GB	35.3 GB	17.3 GB	4-bit · i-quant non-linear — Importance-weighted 4-bit, non-linear	Fits: 32K context · ~34 tok/s
20.4 GB	23.8 GB	36.5 GB	18.3 GB	4-bit medium — 4-bit medium quantization — most popular sweet spot	Fits: 32K context · ~32 tok/s
20.8 GB	24.2 GB	36.9 GB	19.9 GB	4-bit large — 4-bit large quantization	Fits: 16K context · ~32 tok/s
21.2 GB	24.6 GB	37.3 GB	21.3 GB	5-bit · legacy — 5-bit legacy quantization	Fits: 16K context · ~31 tok/s
23.1 GB	26.6 GB	39.2 GB	21.1 GB	5-bit small — 5-bit small quantization	Fits: 8K context · ~28 tok/s
23.9 GB	27.3 GB	40 GB	21.7 GB	5-bit medium — 5-bit medium quantization — good quality/size tradeoff	Fits: 4K context · ~27 tok/s
24.3 GB	27.7 GB	40.4 GB	23 GB	5-bit large — 5-bit large quantization	Does not fit: needs 25.6 GB
27.4 GB	30.9 GB	43.5 GB	25.2 GB	6-bit, very good — 6-bit quantization, very good quality	Does not fit: needs 28.7 GB
32.9 GB	36.3 GB	49 GB	32.6 GB	8-bit, near-lossless — 8-bit quantization, near-lossless	Does not fit: needs 34.2 GB
64.2 GB	67.6 GB	80.3 GB	61.4 GB	16-bit · bfloat — Brain floating point 16 — preferred for training	Does not fit: needs 65.5 GB

Selected Q4_K_M

4-bit medium quantization

21.7GBat 16K context

Weights: 18.8 GB
KV cache: 2.6 GB
Overhead: 0.3 GB

Fits RTX 4090 (reference) · 2.3 GB free · ~32 tok/s

≥50 tok/s needs ≥1.57 TB/s · RTX 5090 ~57

Which Hardware Can Run Gemma 4 31B IT?

Q4_K_M · 20.4 GB

Gemma 4 31B IT (Q4_K_M) requires 20.4 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 27+ GB is recommended. Using the full 262K context window can add up to 33.0 GB, bringing total usage to 53.4 GB. 7 GPUs can run it, including NVIDIA GeForce RTX 5090, NVIDIA GeForce RTX 3090 Ti. 122 devices with unified memory can run Gemma 4 31B IT, including Mac Studio (M3 Ultra, 256GB), Mac Studio (M3 Ultra, 512GB), Mac Studio M1 Max (32 GB).

By VRAM tier · typical GPUs and Macs at Q4_K_M

Typical hardwareQuantVRAMtok/sContext that fits

8 GB · RTX 4060Doesn't fit — smallest quant needs 11.5 GB8 GB tier →

12 GB · RTX 3060 12GBIQ2_XXS · 11.5 GB · ~23 tok/s · 16K11.5~2316K12 GB tier →

16 GB · RTX 5060 Ti 16GBIQ3_XS · 15.8 GB · ~20 tok/s · 16K15.8~2016K16 GB tier →

24 GB · RTX 4090referenceQ4_K_M · 21.7 GB · ~32 tok/s · 32K21.7~3232K24 GB tier →

Mac 32 GB · Mac Mini M4Q4_K_M · 21.7 GB · ~4 tok/s · 64K21.7~464KMacs →

Mac 64 GB · Mac Studio M4 MaxQ4_K_M · 21.7 GB · ~19 tok/s · 256K21.7~19256KMacs →

How good is Gemma 4 31B IT?

The llmrun Score merges 14 public benchmark boards into a single figure, where 100 represents the average for open models.

Score vs VRAM in the 24 GB tier

Gemma 4 31B IT
Other models
Best Score for its VRAM
x: VRAM (GB) · y: llmrun Score

#4 of 36 that fit 24 GB

Each dot is a model at the best quant that fits a 24 GB GPU. The line steps up to the best Score available at each amount of VRAM; a dot below it is beaten by a model that needs no more memory.

Full 24 GB ranking →

Benchmarks

GPQA Diamond
Epoch AI · Apr 2026Independent
75.8%
#30/100 open
Independent
LMArena Text
LMArena · Oct 2026Arena
1452.3
#14/188 open
Arena
AIME 2024/2025
Epoch AI · Apr 2026Independent
73.3%
#28/88 open
Independent
LMArena WebDev
LMArena · Oct 2026Arena
1368.7
#35/55 open
Arena
SciCode
Epoch AI · Apr 2026Lab-reported
45.5%
#25/72 open
Lab-reported
SimpleQA
Epoch AI · Apr 2026Independent
10.4%
#19/19 open
Independent

All 20 Gemma 4 31B IT benchmarks

In the 24 GB class

36 models fit 24 GB

Same Score, only models that fit 24 GB at their best quant: the ranking a 24 GB owner actually chooses from.

#1
Qwen3.8 27BAlibaba
122.0
~38 tok/s · 17.9 GB
#2
Qwen3.6 27BAlibaba
111.2
~38 tok/s · 17.9 GB
#3
Qwen3.6 35B A3BAlibaba
106.8
~151 tok/s · 22.0 GB
#4
Gemma 4 31B ITGoogle
106.7
~32 tok/s · 21.7 GB
#5
Gemma 4 26B A4B ITGoogle
105.8
~171 tok/s · 16.5 GB
#6
Qwen3.5 35B A3BAlibaba
105.8
~151 tok/s · 22.0 GB

Full 24 GB ranking →RTX 4090 page →

Alternatives to Gemma 4 31B IT

Gemma 4 family · 35 models →

Official models with an llmrun Score, each answering one question; numbers are at the best quant that fits a 24 GB GPU.

Best alternative at this size
Qwen3.8 27B
Alibaba
- 15.3 higher Score
- 3.8 GB less VRAM
- 6 tok/s faster
Score 122.0 · 17.9 GB · ~38 tok/s
Compare with Gemma 4 31B IT Open Qwen3.8 27B
Much faster, similar quality
Qwen3.6 35B A3B
Alibaba
- 0.1 higher Score
- 0.3 GB more VRAM
- 119 tok/s faster
Score 106.8 · 22 GB · ~151 tok/s
Compare with Gemma 4 31B IT Open Qwen3.6 35B A3B
Fits in less VRAM
GPT OSS 20B
OpenAI
- 10.3 lower Score
- 8.6 GB less VRAM
- 158 tok/s faster
Score 96.4 · 13.1 GB · ~190 tok/s
Compare with Gemma 4 31B IT Open GPT OSS 20B
Other vendor
Qwen3.6 27B
Alibaba
- 4.5 higher Score
- 3.8 GB less VRAM
- 6 tok/s faster
Score 111.2 · 17.9 GB · ~38 tok/s
Compare with Gemma 4 31B IT Open Qwen3.6 27B

Similar models

10tok/s on RTX 4090

Model	Score	VRAM	tok/s	Why it is similar
Gemma 4 26B A4B ITGoogle	105.8	16.5 GB	~171	similar score
Qwen3.5 35B A3BAlibaba	105.8	22 GB	~151	similar score
Qwen3.5 27BAlibaba	103.8	17.9 GB	~38	similar score
Seed OSS 36B InstructByteDance-Seed	102.5	22.8 GB	~31	similar score
Qwen3 30B A3B Thinking 2507Alibaba	100.7	19.4 GB	~173	similar score
Qwen3.5 9BAlibaba	99.4	6.6 GB	~103	similar score
Qwen3 32BAlibaba	98.6	22.6 GB	~32	similar score
Qwen3.5 4BAlibaba	98.5	3.4 GB	~201	similar score
Qwen3 14BAlibaba	98.0	11.9 GB	~69	similar score
Qwen3 8BAlibaba	93.9	7.6 GB	~119	similar score

Fine-tunes & variants(10)community builds of Gemma 4 31B IT

Specifications

Publisher: Google
Family: Gemma 4
Parameters: 31.3B
Architecture: Gemma4ForConditionalGeneration
Context Length: 262,144 tokens
Vocabulary Size: 262,144
Release Date: 2026-03-11
License: Apache 2.0

Resources

Frequently Asked Questions

How much VRAM does Gemma 4 31B IT need?

Gemma 4 31B IT requires 20.4 GB of VRAM at Q4_K_M, or 64.2 GB at BF16. Full 262K context adds up to 33.0 GB (53.4 GB total).

VRAM = Weights + KV Cache + Overhead

Weights = 31.3B × 4.8 bits ÷ 8 = 18.8 GB

KV Cache + Overhead ≈ 1.6 GB (at 2K context + ~0.3 GB framework)

Fit ratings and hardware model lists check this model with room for a 16K-token context, which needs a little more memory.

KV Cache + Overhead ≈ 34.6 GB (at full 262K context)

VRAM usage by quantization

Q4_K_M

20.4 GB

Q4_K_M + full context

53.4 GB

Learn more about VRAM estimation →

Can NVIDIA GeForce RTX 4090 run Gemma 4 31B IT?

Yes, at Q5_K_M (23.9 GB) or lower. Higher quantizations like Q5_K_L (24.3 GB) exceed the NVIDIA GeForce RTX 4090's 24 GB.

What's the best quantization for Gemma 4 31B IT?

For Gemma 4 31B IT, Q4_K_M (20.4 GB) offers the best balance of quality and VRAM usage. Q4_K_L (20.8 GB) provides better quality if you have the VRAM. The smallest option is IQ2_XXS at 10.2 GB.

VRAM requirement by quantization

IQ2_XXS

10.2 GB~53%

IQ3_S

14.9 GB~75%

Q3_K_L

17.6 GB~86%

Q4_K_M ★

20.4 GB~89%

Q5_0

21.2 GB~90%

BF16

64.2 GB~100%

★ Recommended — best balance of quality and VRAM usage.

Learn more about quantization →

Can I run Gemma 4 31B IT on a Mac?

Yes — Mac Mini M2 Pro (16 GB) and 133 other Macs can run Gemma 4 31B IT, the smallest at IQ2_XS. Apple Silicon uses unified memory, so the model shares RAM with the system. At Q4_K_M you need at least 20.4 GB of usable unified memory (RAM minus macOS overhead), from Mac Mini M5 Pro (24 GB).

Can I run Gemma 4 31B IT locally?

Yes — Gemma 4 31B IT can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.

How fast is Gemma 4 31B IT?

At Q4_K_M, Gemma 4 31B IT can reach ~235 tok/s on AMD Instinct MI350X. On NVIDIA GeForce RTX 4090: ~32 tok/s. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.

tok/s = (bandwidth GB/s ÷ model GB) × efficiency

Example: NVIDIA B200 → 8000 ÷ 20.4 × 0.65 = ~255 tok/s

Estimated speed at Q4_K_M (20.4 GB)

NVIDIA B200

~255 tok/s

NVIDIA GeForce RTX 4090

~32 tok/s

NVIDIA B300

~255 tok/s

AMD Instinct MI350X

~235 tok/s

Real-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.

Learn more about tok/s estimation →

What's the download size of Gemma 4 31B IT?

At Q4_K_M, the download is about 18.76 GB. The full-precision BF16 version is 62.55 GB. The smallest option (IQ2_XXS) is 8.60 GB.

Which GPUs can run Gemma 4 31B IT?

7 consumer GPUs can run Gemma 4 31B IT at Q4_K_M. Top options include NVIDIA GeForce RTX 5090, AMD Radeon RX 7900 XTX, NVIDIA GeForce RTX 3090. 1 GPU has plenty of headroom for comfortable inference.

Which devices can run Gemma 4 31B IT?

122 devices with unified memory can run Gemma 4 31B IT at Q4_K_M (20.4 GB), including AMD Ryzen AI 9 HX 370 (Strix Point) Laptop, ASUS Ascent GX10, Beelink GTR9 Pro (Ryzen AI Max+ 395, 128 GB), Framework Desktop (Ryzen AI Max+ 395, 128 GB). Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.

Gemma 4 31B IT — Hardware Requirements & GPU Compatibility

Decent on a 24 GB GPU

llmrun Composite Score →

Run Gemma 4 31B IT

Recommended

Ways to runOther ways

Run in cloud

How Much VRAM Does Gemma 4 31B IT Need?

Selected Q4_K_M

Which Hardware Can Run Gemma 4 31B IT?

How good is Gemma 4 31B IT?

Score vs VRAM in the 24 GB tier

Benchmarks

In the 24 GB class

Alternatives to Gemma 4 31B IT

Qwen3.8 27B

Qwen3.6 35B A3B

GPT OSS 20B

Qwen3.6 27B

Similar models

Specifications

Resources

Frequently Asked Questions