Skip to content

Qwen3 14B VRAM requirements

How much VRAM Qwen3 14B needs
4-bit weights: 8.3 GBWith 8K context: 10.5 GBWith 32K context: 14.2 GB8-bit, 8K context: 16.8 GB
Parameters
14.8B
Longest context
40K
Released
Apr 2025
License
Apache 2.0
Strongest at
Not rated

How much VRAM Qwen3 14B needs

Memory Qwen3 14B needs at each precision and context length
PrecisionWeightsWith 8KWith 32KWith 40K
16-bit27.5 GB29.7 GB33.4 GB34.7 GB
8-bit14.6 GB16.8 GB20.5 GB21.8 GB
6-bit11.3 GB13.4 GB17.2 GB18.4 GB
5-bit9.8 GB11.9 GB15.7 GB16.9 GB
4-bit8.3 GB10.5 GB14.2 GB15.5 GB
3-bit6.7 GB8.9 GB12.6 GB13.9 GB
2-bit5.2 GB7.3 GB11.1 GB12.3 GB
1.6-bit2.8 GB5.0 GB8.7 GB10.0 GB

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

GPUs and Macs that run Qwen3 14B

The best way Qwen3 14B runs on common GPUs, Macs and AI boxes
MachineBest fitContextSpeedFirst replyReport
RTX 4060Too bigPDF: Qwen3 14B on RTX 4060 report
RTX 5060Too bigPDF: Qwen3 14B on RTX 5060 report
RTX 3070Too bigPDF: Qwen3 14B on RTX 3070 report
RTX 3080 10 GBToo bigPDF: Qwen3 14B on RTX 3080 10 GB report
RTX 3060 12 GB3-bit24K35 tok/s2.3 sPDF: Qwen3 14B on RTX 3060 12 GB report
Arc B5803-bit24K31 tok/s1.6 sPDF: Qwen3 14B on Arc B580 report
RTX 40703-bit24K49 tok/s1.0 sPDF: Qwen3 14B on RTX 4070 report
RTX 4070 Super3-bit24K49 tok/s0.8 sPDF: Qwen3 14B on RTX 4070 Super report
RTX 50703-bit24K64 tok/s1.0 sPDF: Qwen3 14B on RTX 5070 report
RTX 4060 Ti 16 GB6-bit16K17 tok/s1.3 sPDF: Qwen3 14B on RTX 4060 Ti 16 GB report
RTX 5060 Ti 16 GB6-bit16K26 tok/s1.2 sPDF: Qwen3 14B on RTX 5060 Ti 16 GB report
Arc A770 16 GB6-bit16K23 tok/s1.4 sPDF: Qwen3 14B on Arc A770 16 GB report
Radeon RX 9070 XT6-bit16K31 tok/s0.8 sPDF: Qwen3 14B on Radeon RX 9070 XT report
RTX 4070 Ti Super6-bit16K39 tok/s0.7 sPDF: Qwen3 14B on RTX 4070 Ti Super report
RTX 4080 Super6-bit16K43 tok/s0.6 sPDF: Qwen3 14B on RTX 4080 Super report
RTX 5070 Ti6-bit16K51 tok/s0.7 sPDF: Qwen3 14B on RTX 5070 Ti report
RTX 50806-bit16K55 tok/s0.5 sPDF: Qwen3 14B on RTX 5080 report
Radeon RX 7900 XTX8-bit40K36 tok/s1.3 sPDF: Qwen3 14B on Radeon RX 7900 XTX report
RTX 30908-bit40K42 tok/s0.8 sPDF: Qwen3 14B on RTX 3090 report
RTX 40908-bit40K45 tok/s0.4 sPDF: Qwen3 14B on RTX 4090 report
Radeon AI PRO R97008-bit40K25 tok/s0.8 sPDF: Qwen3 14B on Radeon AI PRO R9700 report
RTX 50908-bit40K77 tok/s0.3 sPDF: Qwen3 14B on RTX 5090 report
2× RTX 30908-bit40K42 tok/s0.5 sPDF: Qwen3 14B on 2× RTX 3090 report
2× RTX 40908-bit40K45 tok/s0.2 sPDF: Qwen3 14B on 2× RTX 4090 report
RTX PRO 5000 Blackwell 48 GB8-bit40K59 tok/s0.4 sPDF: Qwen3 14B on RTX PRO 5000 Blackwell 48 GB report
RTX 6000 Ada8-bit40K43 tok/s0.3 sPDF: Qwen3 14B on RTX 6000 Ada report
L40S8-bit40K39 tok/s0.5 sPDF: Qwen3 14B on L40S report
2× RTX 50908-bit40K77 tok/s0.2 sPDF: Qwen3 14B on 2× RTX 5090 report
4× RTX 30908-bit40K42 tok/s0.3 sPDF: Qwen3 14B on 4× RTX 3090 report
RTX PRO 6000 Blackwell8-bit40K77 tok/s0.2 sPDF: Qwen3 14B on RTX PRO 6000 Blackwell report
Mac mini M4 16 GB3-bit16K14 tok/s19 sPDF: Qwen3 14B on Mac mini M4 16 GB report
Mac mini M4 24 GB6-bit20K8.3 tok/s19 sPDF: Qwen3 14B on Mac mini M4 24 GB report
Mac mini M4 32 GB8-bit28K6.5 tok/s19 sPDF: Qwen3 14B on Mac mini M4 32 GB report
Mac mini M4 Pro 24 GB6-bit20K18 tok/s9.3 sPDF: Qwen3 14B on Mac mini M4 Pro 24 GB report
Mac mini M4 Pro 48 GB8-bit40K14 tok/s9.3 sPDF: Qwen3 14B on Mac mini M4 Pro 48 GB report
Mac mini M4 Pro 64 GB8-bit40K14 tok/s9.3 sPDF: Qwen3 14B on Mac mini M4 Pro 64 GB report
M4 Max 128 GB8-bit40K25 tok/s4.7 sPDF: Qwen3 14B on M4 Max 128 GB report
M5 Max 128 GB8-bit40K28 tok/s1.4 sPDF: Qwen3 14B on M5 Max 128 GB report
M3 Ultra 96 GB8-bit40K32 tok/s2.6 sPDF: Qwen3 14B on M3 Ultra 96 GB report
M3 Ultra 256 GB8-bit40K32 tok/s2.6 sPDF: Qwen3 14B on M3 Ultra 256 GB report
M3 Ultra 512 GB8-bit40K32 tok/s2.6 sPDF: Qwen3 14B on M3 Ultra 512 GB report
M5 Ultra 512 GB8-bit40K45 tok/s0.8 sPDF: Qwen3 14B on M5 Ultra 512 GB report
Ryzen AI Max+ 395 64 GB8-bit40K10 tok/s2.7 sPDF: Qwen3 14B on Ryzen AI Max+ 395 64 GB report
Ryzen AI Max+ 395 128 GB8-bit40K10 tok/s2.7 sPDF: Qwen3 14B on Ryzen AI Max+ 395 128 GB report
RTX Spark8-bit40K14 tok/s0.9 sPDF: Qwen3 14B on RTX Spark report
DGX Spark8-bit40K13 tok/s0.9 sPDF: Qwen3 14B on DGX Spark report
Dual DGX Spark8-bit40K13 tok/s0.5 sPDF: Qwen3 14B on Dual DGX Spark report
DGX Station GB3008-bit40K239 tok/s0.1 sPDF: Qwen3 14B on DGX Station GB300 report
2× RTX PRO 6000 Blackwell8-bit40K77 tok/s0.1 sPDF: Qwen3 14B on 2× RTX PRO 6000 Blackwell report
4× RTX PRO 6000 Blackwell8-bit40K77 tok/s0.1 sPDF: Qwen3 14B on 4× RTX PRO 6000 Blackwell report
8× RTX PRO 6000 Blackwell8-bit40K77 tok/s0.1 sPDF: Qwen3 14B on 8× RTX PRO 6000 Blackwell report
A100 80 GB SXM8-bit40K86 tok/s0.6 sPDF: Qwen3 14B on A100 80 GB SXM report
H100 SXM8-bit40K133 tok/s0.2 sPDF: Qwen3 14B on H100 SXM report
4× H200 NVL8-bit40K178 tok/s0.1 sPDF: Qwen3 14B on 4× H200 NVL report
8× H200 NVL8-bit40K178 tok/s0.1 sPDF: Qwen3 14B 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/Qwen3-14B-GGUF
vLLM
vllm serve Qwen/Qwen3-14B
MLX (Mac)
mlx_lm.chat --model mlx-community/Qwen3-14B-4bit

Questions

How much VRAM does Qwen3 14B need?

Qwen3 14B needs 8.3 GB for its weights at 4-bit, 10.5 GB with an 8K context and 14.2 GB with 32K. At 8-bit, 16.8 GB.

What is the minimum VRAM for Qwen3 14B?

About 8.9 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 RTX 3060 12 GB.

How big is Qwen3 14B?

14.8B parameters: 27.5 GB of weights at 16-bit, 8.3 GB at 4-bit.

Can Qwen3 14B run on an RTX 4090?

Yes. At 8-bit it holds 40K of context and writes about 45 tokens a second.

More models