AlibabaQwenQWenLMHeadModel

Qwen 14B — Hardware Requirements & GPU Compatibility

Chat

Qwen-14B is Alibaba's first-generation 14-billion-parameter base language model, pretrained from scratch on over 3 trillion tokens of Chinese, English, multilingual, code, and math data, with a roughly 150,000-token vocabulary for broader multilingual coverage. It is a raw pretrained Transformer, not tuned for conversation; Alibaba's aligned Qwen-14B-Chat is the assistant built on top of it. The card reports it beating other open models of similar size, and in some benchmarks larger models too, on Chinese and English evaluation suites. At 14B parameters it needs a capable consumer GPU at full precision, considerably less once quantized. Context length is 8,192 tokens, extendable further with the NTK-aware interpolation and window-attention techniques described in the card. It is released under a custom Tongyi Qianwen License Agreement that is free for research, with commercial use requiring a separate application to Alibaba. It was published in September 2023.

14.2B params · dense8K contextSep 20234.0K downloads221 likes868 quant downloads

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.

S

Runs great on a 24 GB GPU

Runs from 8 GB (IQ3_S · 8K)Best on 24 GB (Q4_K_M · ~70 tok/s · 8K)Macs from 16 GB (Q4_K_M · ~15 tok/s)

Reference GPU · RTX 4090 24 GB

Estimates, ±20% · how we estimate

Run Qwen 14B

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.

Recommended

Ollama

one command · needs Ollama installed

$ollama run hf.co/mradermacher/Qwen-14B-i1-GGUF:Q4_K_M

Q4_K_M fits fully on the RTX 4090 · 13.4 GB free · 8K context

Download: Qwen-14B.i1-Q4_K_M.gguf9.45 GBmradermacher

install, OS, other quants

Ways to runOther ways

each opens step-by-step

Run in cloud

Runs on RTX 5070 Ti (16 GB) (5 GB headroom) · Q4_K_M

Generation speed: ~62 tok/s; generation speed
Cost per 1M output tokens: $0.75; per 1M output tokens

Deployon Vast.aiBest valuefrom $0.17/hr

Compare GPUs →

or

How Much VRAM Does Qwen 14B Need?

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

Quantization	VRAM	+32K	+128K	File	Quality	On RTX 4090 (reference)
	4.3 GB	—	—	4.9 GB	2-bit · i-quant extreme compression — Importance-weighted 2-bit, extreme compression — significant quality loss	Fits: 8K context · ~153 tok/s
	4.7 GB	—	—	5.2 GB	2-bit · i-quant extra small — Importance-weighted 2-bit, extra small	Fits: 8K context · ~140 tok/s
	4.9 GB	—	—	5.4 GB	2-bit · i-quant small — Importance-weighted 2-bit, small	Fits: 8K context · ~135 tok/s
	5.3 GB	—	—	5.7 GB	2-bit · i-quant medium — Importance-weighted 2-bit, medium	Fits: 8K context · ~125 tok/s
	6.0 GB	—	—	6.3 GB	3-bit · i-quant — Importance-weighted 3-bit	Fits: 8K context · ~109 tok/s
	6.2 GB	—	—	5.7 GB	2-bit · small K-quant — 2-bit small K-quant	Fits: 8K context · ~105 tok/s
	6.4 GB	—	—	6.6 GB	3-bit · i-quant extra small — Importance-weighted 3-bit, extra small	Fits: 8K context · ~102 tok/s
	6.6 GB	—	—	6.8 GB	3-bit · i-quant small — Importance-weighted 3-bit, small	Fits: 8K context · ~99 tok/s
	6.6 GB	—	—	5.8 GB	2-bit · K-quant — 2-bit quantization with K-quant improvements	Fits: 8K context · ~99 tok/s
	6.8 GB	—	—	6.8 GB	3-bit small — 3-bit small quantization	Fits: 8K context · ~96 tok/s
	7.0 GB	—	—	7.4 GB	3-bit · i-quant medium — Importance-weighted 3-bit, medium	Fits: 8K context · ~94 tok/s
	7.6 GB	—	—	7.7 GB	3-bit medium — 3-bit medium quantization	Fits: 8K context · ~86 tok/s
	7.8 GB	—	—	8.2 GB	4-bit · legacy — 4-bit legacy quantization	Fits: 8K context · ~84 tok/s
	8.0 GB	—	—	8 GB	3-bit large — 3-bit large quantization	Fits: 8K context · ~82 tok/s
	8.4 GB	—	—	7.8 GB	4-bit · i-quant compact — Importance-weighted 4-bit, compact	Fits: 8K context · ~78 tok/s
	8.8 GB	—	—	9 GB	4-bit · legacy+ — 4-bit legacy quantization with offset	Fits: 8K context · ~75 tok/s
	8.8 GB	—	—	8.6 GB	4-bit small — 4-bit small quantization	Fits: 8K context · ~75 tok/s
	8.8 GB	—	—	8.2 GB	4-bit · i-quant non-linear — Importance-weighted 4-bit, non-linear	Fits: 8K context · ~75 tok/s
	9.3 GB	—	—	9.4 GB	4-bit medium — 4-bit medium quantization — most popular sweet spot	Fits: 8K context · ~70 tok/s
	10.7 GB	—	—	10 GB	5-bit small — 5-bit small quantization	Fits: 8K context · ~61 tok/s
	11.1 GB	—	—	10.9 GB	5-bit medium — 5-bit medium quantization — good quality/size tradeoff	Fits: 8K context · ~59 tok/s
	12.9 GB	—	—	12.3 GB	6-bit, very good — 6-bit quantization, very good quality	Fits: 8K context · ~51 tok/s
	15.6 GB	—	—	15.1 GB	8-bit, near-lossless — 8-bit quantization, near-lossless	Fits: 8K context · ~42 tok/s
	31.2 GB	—	—	~28.3 GB	16-bit · bfloat — Brain floating point 16 — preferred for training	Does not fit: needs 31.2 GB

est.calculated, no published file yet

Selected Q4_K_M

4-bit medium quantization

10.6GBat 8K context

Weights: 8.6 GB
KV cache: 1.7 GB
Overhead: 0.3 GB

Fits RTX 4090 (reference) · 13.4 GB free · ~70 tok/s

≥50 tok/s needs ≥719 GB/s · RTX 5090 ~125, RTX 3090 Ti ~70

Which Hardware Can Run Qwen 14B?

Q4_K_M · 9.3 GB

Qwen 14B (Q4_K_M) requires 9.3 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 13+ GB is recommended. 40 GPUs can run it, including NVIDIA GeForce RTX 5090, NVIDIA GeForce RTX 3090 Ti, NVIDIA GeForce RTX 3080 Ti. 150 devices with unified memory can run Qwen 14B, including Mac Studio (M3 Ultra, 256GB), Mac Studio (M3 Ultra, 512GB), Mac Mini M2 Pro (16 GB).

By VRAM tier · typical GPUs and Macs at Q4_K_M

Typical hardwareQuantVRAMtok/sContext that fits

8 GB · RTX 4060IQ3_S · 7.9 GB · ~27 tok/s · 8K7.9~278K8 GB tier →

12 GB · RTX 3060 12GBQ4_K_M · 10.6 GB · ~25 tok/s · 8K10.6~258K12 GB tier →

16 GB · RTX 5060 Ti 16GBQ4_K_M · 10.6 GB · ~31 tok/s · 8K10.6~318K16 GB tier →

24 GB · RTX 4090referenceQ4_K_M · 10.6 GB · ~70 tok/s · 8K10.6~708K24 GB tier →

Mac 32 GB · Mac Mini M4Q4_K_M · 10.6 GB · ~9 tok/s · 8K10.6~98KMacs →

Mac 64 GB · Mac Studio M4 MaxQ4_K_M · 10.6 GB · ~41 tok/s · 8K10.6~418KMacs →

How good is Qwen 14B?

llmrun Score blends public benchmark boards into one number; 100 = the average open model.

Not enough data yet for an llmrun Score.

A Score needs results on enough independent boards. How the Composite Score is built

Benchmarks

MMLU
Epoch AI · Sep 2023Lab-reported
66.3%
#36/85 open
Lab-reported
BIG-Bench Hard
Epoch AI · Sep 2023Lab-reported
53.4%
#18/38 open
Lab-reported
GSM8K
Epoch AI · Sep 2023Lab-reported
61.3%
#24/62 open
Lab-reported
ARC Challenge
Epoch AI · Sep 2023Lab-reported
84.4%
#10/56 open
Lab-reported
BoolQ
Epoch AI · Sep 2023Lab-reported
86.2%
#6/38 open
Lab-reported
PIQA
Epoch AI · Sep 2023Lab-reported
79.9%
#21/40 open
Lab-reported

All 6 Qwen 14B benchmarks

Alternatives to Qwen 14B

Qwen family · 28 models →

Similar models

6tok/s on RTX 4090

Model	Score	VRAM	tok/s	Why it is similar
Apriel 1.6 15B ThinkerServiceNow-AI	—	13.2 GB	~67	same size
Phi 4 ReasoningMicrosoft	—	12.5 GB	~69	same size
Phi 3 Medium 128k InstructMicrosoft	—	12 GB	~72	same size
JiRackUltra 14BCMSManhattan	—	12.4 GB	~69	same size
Nemotron Terminal 14BNVIDIA	—	11.9 GB	~69	same size
Vllm Translategemma 12B ITchbae624	—	9.8 GB	~73	same size

Specifications

Publisher: Alibaba
Family: Qwen
Parameters: 14.2B
Architecture: QWenLMHeadModel
Context Length: 8,192 tokens
Vocabulary Size: 152,064
Release Date: 2023-09-24

Resources

Frequently Asked Questions

How much VRAM does Qwen 14B need?

Qwen 14B requires 9.3 GB of VRAM at Q4_K_M, or 31.2 GB at BF16.

VRAM = Weights + KV Cache + Overhead

Weights = 14.2B × 4.8 bits ÷ 8 = 8.5 GB

KV Cache + Overhead ≈ 0.8 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.

VRAM usage by quantization

Q4_K_M

9.3 GB

Learn more about VRAM estimation →

Can NVIDIA GeForce RTX 4090 run Qwen 14B?

Yes, at Q8_0 (15.6 GB) or lower. Higher quantizations like BF16 (31.2 GB) exceed the NVIDIA GeForce RTX 4090's 24 GB.

What's the best quantization for Qwen 14B?

For Qwen 14B, Q4_K_M (9.3 GB) offers the best balance of quality and VRAM usage. Q5_K_S (10.7 GB) provides better quality if you have the VRAM. The smallest option is IQ2_XXS at 4.3 GB.

VRAM requirement by quantization

IQ2_XXS

4.3 GB~53%

IQ3_XS

6.4 GB~73%

Q4_0

7.8 GB~85%

IQ4_NL

8.8 GB~88%

Q4_K_M ★

9.3 GB~89%

BF16

31.2 GB~100%

★ Recommended — best balance of quality and VRAM usage.

Learn more about quantization →

Can I run Qwen 14B on a Mac?

Yes — Mac Mini M2 Pro (16 GB) and 133 other Macs can run Qwen 14B. Apple Silicon uses unified memory, so the model shares RAM with the system. At Q4_K_M you need at least 9.3 GB of usable unified memory (RAM minus macOS overhead).

Can I run Qwen 14B locally?

Yes — Qwen 14B can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.

How fast is Qwen 14B?

At Q4_K_M, Qwen 14B can reach ~513 tok/s on AMD Instinct MI350X. On NVIDIA GeForce RTX 4090: ~70 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 ÷ 9.35 × 0.65 = ~556 tok/s

Estimated speed at Q4_K_M (9.3 GB)

NVIDIA B200

~556 tok/s

NVIDIA GeForce RTX 4090

~70 tok/s

NVIDIA B300

~556 tok/s

AMD Instinct MI350X

~513 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 Qwen 14B?

At Q4_K_M, the download is about 8.50 GB. The full-precision BF16 version is 28.33 GB. The smallest option (IQ2_XXS) is 3.90 GB.

Which GPUs can run Qwen 14B?

40 consumer GPUs can run Qwen 14B at Q4_K_M. Top options include AMD Radeon RX 6800, AMD Radeon RX 6800 XT, AMD Radeon RX 6900 XT, AMD Radeon RX 6700 XT. 26 GPUs have plenty of headroom for comfortable inference.

Which devices can run Qwen 14B?

150 devices with unified memory can run Qwen 14B at Q4_K_M (9.3 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.

Run Qwen 14B

10.6 GB VRAM at 8K context · fits the RTX 4090 (13.4 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/mradermacher/Qwen-14B-i1-GGUF:Q4_K_M
Raise the context window
Type it inside the chat. For the API server instead, start it with OLLAMA_CONTEXT_LENGTH=8192 ollama serve.
$/set parameter num_ctx 8192
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 paste https://huggingface.co/mradermacher/Qwen-14B-i1-GGUF, then pick Q4_K_M. Or use the CLI (run LM Studio once first):
$lms get https://huggingface.co/mradermacher/Qwen-14B-i1-GGUF@Q4_K_M
Open on Hugging Face ↗
Load and chat
Open the model loader, select the model and set Context Length to 8192. 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 mradermacher/Qwen-14B-i1-GGUF:Q4_K_M -c 8192 -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/mradermacher/Qwen-14B-i1-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 mradermacher/Qwen-14B-i1-GGUF:Q4_K_M -c 8192 -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 Qwen/Qwen-14B --max-model-len 8192
Runs the original weights, not the Q4_K_M GGUF; needs ~31 GB of GPU memory.

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.
$mradermacher/Qwen-14B-i1-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 5070 Ti (16 GB)CheapestCheapest5 GB free	~62 tok/s	$0.75	Deploy on Vast.aifrom $0.17/hr (opens in new tab)	16 GB · 5 GB free · ~62 tok/s · $0.75 / 1M tokens
RTX A4000 (16 GB)5 GB free	~31 tok/s	$0.79	Deploy on Vast.aifrom $0.088/hr (opens in new tab)	16 GB · 5 GB free · ~31 tok/s · $0.79 / 1M tokens
RTX 5060 Ti 16GB5 GB free	~31 tok/s	$0.99	Deploy on Vast.aifrom $0.11/hr (opens in new tab)	5 GB free · ~31 tok/s · $0.99 / 1M tokens
T4 (16 GB)5 GB free	~22 tok/s	$1.68	Deploy on Vast.aifrom $0.13/hr (opens in new tab)	16 GB · 5 GB free · ~22 tok/s · $1.68 / 1M tokens
RTX PRO 4500 (32 GB)21 GB free	~62 tok/s	$1.52	Deploy on RunPodfrom $0.34/hr (opens in new tab)	32 GB · 21 GB free · ~62 tok/s · $1.52 / 1M tokens
RTX 4090 (24 GB)13 GB free	~70 tok/s	$1.35	Deploy on RunPodfrom $0.34/hr (opens in new tab)	24 GB · 13 GB free · ~70 tok/s · $1.35 / 1M tokens
A40 (48 GB)37 GB free	~48 tok/s	$2.01	Deploy on RunPodfrom $0.35/hr (opens in new tab)	48 GB · 37 GB free · ~48 tok/s · $2.01 / 1M tokens
L4 (24 GB)13 GB free	~21 tok/s	$5.85	Deploy on RunPodfrom $0.44/hr (opens in new tab)	24 GB · 13 GB free · ~21 tok/s · $5.85 / 1M tokens

Compare GPUs →