Skip to content

Qwen3 30B A3B VRAM requirements

How much VRAM Qwen3 30B A3B needs
4-bit weights: 17.2 GBWith 8K context: 18.8 GBWith 32K context: 21.0 GB8-bit, 8K context: 31.8 GB
Parameters
30.5B
Active per token
3B
Longest context
40K
Released
Apr 2025
License
Apache 2.0
Strongest at
Not rated

How much VRAM Qwen3 30B A3B needs

Memory Qwen3 30B A3B needs at each precision and context length
PrecisionWeightsWith 8KWith 32KWith 40K
16-bit56.9 GB58.4 GB60.7 GB61.4 GB
8-bit30.2 GB31.8 GB34.0 GB34.8 GB
6-bit23.3 GB24.9 GB27.1 GB27.9 GB
5-bit20.2 GB21.8 GB24.0 GB24.8 GB
4-bit17.2 GB18.8 GB21.0 GB21.8 GB
3-bit13.9 GB15.5 GB17.7 GB18.5 GB
2-bit10.7 GB12.2 GB14.5 GB15.2 GB
1.6-bit5.9 GB7.4 GB9.7 GB10.4 GB

One conversation, 16-bit context cache, runtime overhead included.

GPUs and Macs that run Qwen3 30B A3B

The best way Qwen3 30B A3B runs on common GPUs, Macs and AI boxes
MachineBest fitContextSpeedFirst replyReport
RTX 4060Too bigPDF: Qwen3 30B A3B on RTX 4060 report
RTX 5060Too bigPDF: Qwen3 30B A3B on RTX 5060 report
RTX 3070Too bigPDF: Qwen3 30B A3B on RTX 3070 report
RTX 3080 10 GBToo bigPDF: Qwen3 30B A3B on RTX 3080 10 GB report
RTX 3060 12 GBToo bigPDF: Qwen3 30B A3B on RTX 3060 12 GB report
Arc B580Too bigPDF: Qwen3 30B A3B on Arc B580 report
RTX 4070Too bigPDF: Qwen3 30B A3B on RTX 4070 report
RTX 4070 SuperToo bigPDF: Qwen3 30B A3B on RTX 4070 Super report
RTX 5070Too bigPDF: Qwen3 30B A3B on RTX 5070 report
RTX 4060 Ti 16 GBToo bigPDF: Qwen3 30B A3B on RTX 4060 Ti 16 GB report
RTX 5060 Ti 16 GBToo bigPDF: Qwen3 30B A3B on RTX 5060 Ti 16 GB report
Arc A770 16 GBToo bigPDF: Qwen3 30B A3B on Arc A770 16 GB report
Radeon RX 9070 XTToo bigPDF: Qwen3 30B A3B on Radeon RX 9070 XT report
RTX 4070 Ti SuperToo bigPDF: Qwen3 30B A3B on RTX 4070 Ti Super report
RTX 4080 SuperToo bigPDF: Qwen3 30B A3B on RTX 4080 Super report
RTX 5070 TiToo bigPDF: Qwen3 30B A3B on RTX 5070 Ti report
RTX 5080Too bigPDF: Qwen3 30B A3B on RTX 5080 report
Radeon RX 7900 XTX5-bit16K125 tok/s0.8 sPDF: Qwen3 30B A3B on Radeon RX 7900 XTX report
RTX 30905-bit16K163 tok/s0.5 sPDF: Qwen3 30B A3B on RTX 3090 report
RTX 40905-bit16K169 tok/s0.2 sPDF: Qwen3 30B A3B on RTX 4090 report
Radeon AI PRO R97006-bit40K93 tok/s0.5 sPDF: Qwen3 30B A3B on Radeon AI PRO R9700 report
RTX 50906-bit40K207 tok/s0.2 sPDF: Qwen3 30B A3B on RTX 5090 report
2× RTX 30908-bit40K129 tok/s0.3 sPDF: Qwen3 30B A3B on 2× RTX 3090 report
2× RTX 40908-bit40K135 tok/s0.1 sPDF: Qwen3 30B A3B on 2× RTX 4090 report
RTX PRO 5000 Blackwell 48 GB8-bit40K160 tok/s0.3 sPDF: Qwen3 30B A3B on RTX PRO 5000 Blackwell 48 GB report
RTX 6000 Ada8-bit40K131 tok/s0.2 sPDF: Qwen3 30B A3B on RTX 6000 Ada report
L40S8-bit40K123 tok/s0.3 sPDF: Qwen3 30B A3B on L40S report
2× RTX 50908-bit40K184 tok/s0.1 sPDF: Qwen3 30B A3B on 2× RTX 5090 report
4× RTX 30908-bit40K129 tok/s0.2 sPDF: Qwen3 30B A3B on 4× RTX 3090 report
RTX PRO 6000 Blackwell8-bit40K184 tok/s0.1 sPDF: Qwen3 30B A3B on RTX PRO 6000 Blackwell report
Mac mini M4 16 GBToo bigPDF: Qwen3 30B A3B on Mac mini M4 16 GB report
Mac mini M4 24 GBToo bigPDF: Qwen3 30B A3B on Mac mini M4 24 GB report
Mac mini M4 32 GB4-bit24K40 tok/s12 sPDF: Qwen3 30B A3B on Mac mini M4 32 GB report
Mac mini M4 Pro 24 GBToo bigPDF: Qwen3 30B A3B on Mac mini M4 Pro 24 GB report
Mac mini M4 Pro 48 GB8-bit32K48 tok/s5.9 sPDF: Qwen3 30B A3B on Mac mini M4 Pro 48 GB report
Mac mini M4 Pro 64 GB8-bit40K48 tok/s5.9 sPDF: Qwen3 30B A3B on Mac mini M4 Pro 64 GB report
M4 Max 128 GB8-bit40K69 tok/s2.9 sPDF: Qwen3 30B A3B on M4 Max 128 GB report
M5 Max 128 GB8-bit40K73 tok/s0.9 sPDF: Qwen3 30B A3B on M5 Max 128 GB report
M3 Ultra 96 GB8-bit40K79 tok/s1.7 sPDF: Qwen3 30B A3B on M3 Ultra 96 GB report
M3 Ultra 256 GB8-bit40K79 tok/s1.7 sPDF: Qwen3 30B A3B on M3 Ultra 256 GB report
M3 Ultra 512 GB8-bit40K79 tok/s1.7 sPDF: Qwen3 30B A3B on M3 Ultra 512 GB report
M5 Ultra 512 GB8-bit40K93 tok/s0.5 sPDF: Qwen3 30B A3B on M5 Ultra 512 GB report
Ryzen AI Max+ 395 64 GB8-bit40K40 tok/s1.7 sPDF: Qwen3 30B A3B on Ryzen AI Max+ 395 64 GB report
Ryzen AI Max+ 395 128 GB8-bit40K40 tok/s1.7 sPDF: Qwen3 30B A3B on Ryzen AI Max+ 395 128 GB report
RTX Spark8-bit40K55 tok/s0.6 sPDF: Qwen3 30B A3B on RTX Spark report
DGX Spark8-bit40K51 tok/s0.7 sPDF: Qwen3 30B A3B on DGX Spark report
Dual DGX Spark8-bit40K51 tok/s0.4 sPDF: Qwen3 30B A3B on Dual DGX Spark report
DGX Station GB3008-bit40K284 tok/s0.1 sPDF: Qwen3 30B A3B on DGX Station GB300 report
2× RTX PRO 6000 Blackwell8-bit40K184 tok/s0.1 sPDF: Qwen3 30B A3B on 2× RTX PRO 6000 Blackwell report
4× RTX PRO 6000 Blackwell8-bit40K184 tok/s0.1 sPDF: Qwen3 30B A3B on 4× RTX PRO 6000 Blackwell report
8× RTX PRO 6000 Blackwell8-bit40K184 tok/s0.1 sPDF: Qwen3 30B A3B on 8× RTX PRO 6000 Blackwell report
A100 80 GB SXM8-bit40K196 tok/s0.4 sPDF: Qwen3 30B A3B on A100 80 GB SXM report
H100 SXM8-bit40K236 tok/s0.1 sPDF: Qwen3 30B A3B on H100 SXM report
4× H200 NVL8-bit40K261 tok/s0.1 sPDF: Qwen3 30B A3B on 4× H200 NVL report
8× H200 NVL8-bit40K261 tok/s0.1 sPDF: Qwen3 30B A3B on 8× H200 NVL report

Best fit is the highest quality, up to 8-bit, that leaves 16K of context. First reply is the wait for a 2K prompt. Speeds are rough, plus or minus 25%.

How to run it

llama.cpp
llama-server -hf Qwen/Qwen3-30B-A3B-GGUF
vLLM
vllm serve Qwen/Qwen3-30B-A3B
MLX (Mac)
mlx_lm.chat --model mlx-community/Qwen3-30B-A3B-4bit

Questions

How much VRAM does Qwen3 30B A3B need?

Qwen3 30B A3B needs 17.2 GB for its weights at 4-bit, 18.8 GB with an 8K context and 21.0 GB with 32K. At 8-bit, 31.8 GB.

What is the minimum VRAM for Qwen3 30B A3B?

About 15.5 GB at 3-bit with an 8K context, the lowest precision we recommend for a model this size. The smallest machine here that runs it without spilling is the Radeon RX 7900 XTX.

How big is Qwen3 30B A3B?

30.5B parameters (3B active per token): 56.9 GB of weights at 16-bit, 17.2 GB at 4-bit.

Can Qwen3 30B A3B run on an RTX 4090?

Yes. At 5-bit it holds 16K of context and writes about 169 tokens a second.

More models