FitMyLLM
Alibaba/Mixture of Experts

AlibabaQwen3 30B A3B

Qwen3 30B A3B — MoE running at 3B speed. Great efficiency-quality tradeoff.

chatreasoningThinkingTool Use
30.5B
Parameters (3B active)
32K
Context length
23
Benchmarks
14
Quantizations
400K
HF downloads
Architecture
MoE
Released
2025-04-28
Layers
48
KV Heads
4
Head Dim
128
Family
qwen

Quantization Options

Context length:
QuantBitsVRAM @ 16KQuality
IQ3_XXS3.25
14.0 GB
12.9 + 1.1 KV
low
IQ3_XS3.5
15.0 GB
13.8 + 1.1 KV
low
Q3_K_S3.64
15.5 GB
14.4 + 1.1 KV
low
IQ3_M3.76
15.9 GB
14.8 + 1.1 KV
low
Q3_K_M4
16.9 GB
15.7 + 1.1 KV
low
Q3_K_L4.3
18.0 GB
16.9 + 1.1 KV
moderate
IQ4_XS4.46
18.6 GB
17.5 + 1.1 KV
moderate
Q4_K_S4.67
19.4 GB
18.3 + 1.1 KV
moderate
Q4_K_M4.89
20.3 GB
19.1 + 1.1 KV
good
Q5_K_S5.57
22.8 GB
21.7 + 1.1 KV
good
Q5_K_M5.7
23.3 GB
22.2 + 1.1 KV
good
Q6_K6.56
26.6 GB
25.5 + 1.1 KV
excellent
Q8_08.5
34.0 GB
32.9 + 1.1 KV
lossless
FP1616
62.6 GB
61.5 + 1.1 KV
lossless

Select your GPU above to see speed estimates and compatibility for each quantization.

Deploying for a team or in production? Size GPUs, cost & scaling in Enterprise →
Ready to run this?Rent by the hour

Rent a GPU and run Qwen3 30B A3B now

Spin up an A100 / H100 / 4090 in ~60s. Pay by the second. Cancel anytime.

Community Ratings

Loading ratings...

Benchmarks (23)

Run this model

Easiest way to get started·Beginners
Docs ↗
curl -fsSL https://ollama.com/install.sh | sh
$ollama run qwen3:30b-a3b-q4_K_M

Downloads and runs automatically. Add --verbose for speed stats.

Setup guide
>_

Auto-setup with fitmyllm CLI

Detects your GPU, recommends the best model, downloads it, and starts chatting — zero config. Benchmarks your speed and contributes anonymous data to improve predictions.

pip install fitmyllmthen run fitmyllmLearn more
Auto-detect GPULive tok/s in chatSpeed benchmarks9 inference engines

GPUs that can run this model

At Q4_K_M quantization. Sorted by minimum VRAM.

NVIDIA RTX A4500
20 GB VRAM • 640 GB/s
NVIDIA
$2000
Apple M4 Pro (24GB)
24 GB VRAM • 273 GB/s
APPLE
$1399
NVIDIA L4 24GB
24 GB VRAM • 300 GB/s
NVIDIA
$2500
Apple M2 (24GB)
24 GB VRAM • 100 GB/s
APPLE
$999
Apple M3 (24GB)
24 GB VRAM • 100 GB/s
APPLE
$999
Apple M4 (24GB)
24 GB VRAM • 120 GB/s
APPLE
$699
NVIDIA Tesla M40 24 GB
24 GB VRAM • 288 GB/s
NVIDIA
NVIDIA Tesla P10
24 GB VRAM • 694 GB/s
NVIDIA
NVIDIA Tesla P40
24 GB VRAM • 347 GB/s
NVIDIA
NVIDIA RTX A5000
24 GB VRAM • 768 GB/s
NVIDIA
$2500
NVIDIA L40 CNX
24 GB VRAM • 864 GB/s
NVIDIA
$5000

Find the best GPU for Qwen3 30B A3B

Build Hardware for Qwen3 30B A3B

Qwen3 30B A3B — MoE running at 3B speed. Great efficiency-quality tradeoff.

Spec sheet

Qwen3 30B A3B — 30.5B MoE.

Specifications
Parameters
30.5B (3B active)
Architecture
Mixture of Experts
Context Length
32K tokens
Capabilities
chat, reasoning
Release Date
2025-04-28
Provider
Alibaba
Family
qwen
VRAM requirements
QuantBPWVRAMQuality
IQ3_XXS3.2512.9 GB82%
IQ3_XS3.513.8 GB84%
Q3_K_S3.6414.4 GB85%
IQ3_M3.7614.8 GB86%
Q3_K_M415.7 GB88%
Q3_K_L4.316.9 GB90%
IQ4_XS4.4617.5 GB92%
Q4_K_S4.6718.3 GB93%
Q4_K_M4.8919.1 GB94%
Q5_K_S5.5721.7 GB96%
Q5_K_M5.722.2 GB96%
Q6_K6.5625.5 GB97%
Q8_08.532.9 GB100%
FP161661.5 GB100%
Benchmark scores
HumanEval86.0
MMLU-PRO58.0
MATH90.0
IFEval80.0
BBH58.3
GPQA48.0
MUSR19.1
BigCodeBench32.3
Arena Elo1476
GPQA Diamond70.7
LiveCodeBench70.7
AIME56.3
MATH-50097.6
HLE9.8
AA Intelligence22.4
AA Coding14.7
AA Math56.3
IFBench31.9
Terminal-Bench Hard6.8
τ²-Bench22.2
SciCode26.4
AA Long Context22.7
EQ-Bench 3917
Run command

Run Qwen3 30B A3B locally with Ollama — needs 19.1 GB VRAM at Q4_K_M:

$ollama run qwen3:30b
Compatible GPUs
30 @ Q4_K_M