Skip to content

Qwen3 VL 8B Instruct VRAM requirements

How much VRAM Qwen3 VL 8B Instruct needs
4-bit weights: 5.0 GBWith 8K context: 6.9 GBWith 32K context: 10.3 GB8-bit, 8K context: 10.7 GB
Parameters
8.8B
Longest context
256K
Released
Oct 2025
License
Apache 2.0
Strongest at
Text and images

How much VRAM Qwen3 VL 8B Instruct needs

Memory Qwen3 VL 8B Instruct needs at each precision and context length
PrecisionWeightsWith 8KWith 32KWith 128K
16-bit16.3 GB18.3 GB21.7 GB35.2 GB
8-bit8.7 GB10.7 GB14.0 GB27.5 GB
6-bit6.7 GB8.7 GB12.1 GB25.6 GB
5-bit5.8 GB7.8 GB11.2 GB24.7 GB
4-bit5.0 GB6.9 GB10.3 GB23.8 GB
3-bit4.0 GB6.0 GB9.4 GB22.9 GB
2-bit3.1 GB5.1 GB8.4 GB21.9 GB
1.6-bit1.7 GB3.7 GB7.0 GB20.5 GB

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

GPUs and Macs that run Qwen3 VL 8B Instruct

The best way Qwen3 VL 8B Instruct runs on common GPUs, Macs and AI boxes
MachineBest fitContextSpeedFirst replyReport
RTX 40603-bit16K45 tok/s1.1 sPDF: Qwen3 VL 8B Instruct on RTX 4060 report
RTX 50603-bit16K73 tok/s0.9 sPDF: Qwen3 VL 8B Instruct on RTX 5060 report
RTX 30703-bit16K73 tok/s0.8 sPDF: Qwen3 VL 8B Instruct on RTX 3070 report
RTX 3080 10 GB5-bit20K84 tok/s0.6 sPDF: Qwen3 VL 8B Instruct on RTX 3080 10 GB report
RTX 3060 12 GB6-bit24K36 tok/s1.3 sPDF: Qwen3 VL 8B Instruct on RTX 3060 12 GB report
Arc B5806-bit24K32 tok/s0.9 sPDF: Qwen3 VL 8B Instruct on Arc B580 report
RTX 40706-bit24K50 tok/s0.6 sPDF: Qwen3 VL 8B Instruct on RTX 4070 report
RTX 4070 Super6-bit24K50 tok/s0.5 sPDF: Qwen3 VL 8B Instruct on RTX 4070 Super report
RTX 50706-bit24K65 tok/s0.5 sPDF: Qwen3 VL 8B Instruct on RTX 5070 report
RTX 4060 Ti 16 GB8-bit40K23 tok/s0.8 sPDF: Qwen3 VL 8B Instruct on RTX 4060 Ti 16 GB report
RTX 5060 Ti 16 GB8-bit40K35 tok/s0.7 sPDF: Qwen3 VL 8B Instruct on RTX 5060 Ti 16 GB report
Arc A770 16 GB8-bit40K30 tok/s0.8 sPDF: Qwen3 VL 8B Instruct on Arc A770 16 GB report
Radeon RX 9070 XT8-bit40K41 tok/s0.5 sPDF: Qwen3 VL 8B Instruct on Radeon RX 9070 XT report
RTX 4070 Ti Super8-bit40K51 tok/s0.4 sPDF: Qwen3 VL 8B Instruct on RTX 4070 Ti Super report
RTX 4080 Super8-bit40K56 tok/s0.3 sPDF: Qwen3 VL 8B Instruct on RTX 4080 Super report
RTX 5070 Ti8-bit40K67 tok/s0.4 sPDF: Qwen3 VL 8B Instruct on RTX 5070 Ti report
RTX 50808-bit40K72 tok/s0.3 sPDF: Qwen3 VL 8B Instruct on RTX 5080 report
Radeon RX 7900 XTX8-bit88K59 tok/s0.7 sPDF: Qwen3 VL 8B Instruct on Radeon RX 7900 XTX report
RTX 30908-bit88K70 tok/s0.5 sPDF: Qwen3 VL 8B Instruct on RTX 3090 report
RTX 40908-bit88K75 tok/s0.2 sPDF: Qwen3 VL 8B Instruct on RTX 4090 report
Radeon AI PRO R97008-bit144K41 tok/s0.5 sPDF: Qwen3 VL 8B Instruct on Radeon AI PRO R9700 report
RTX 50908-bit144K125 tok/s0.2 sPDF: Qwen3 VL 8B Instruct on RTX 5090 report
2× RTX 30908-bit248K70 tok/s0.3 sPDF: Qwen3 VL 8B Instruct on 2× RTX 3090 report
2× RTX 40908-bit248K75 tok/s0.1 sPDF: Qwen3 VL 8B Instruct on 2× RTX 4090 report
RTX PRO 5000 Blackwell 48 GB8-bit256K97 tok/s0.2 sPDF: Qwen3 VL 8B Instruct on RTX PRO 5000 Blackwell 48 GB report
RTX 6000 Ada8-bit256K72 tok/s0.2 sPDF: Qwen3 VL 8B Instruct on RTX 6000 Ada report
L40S8-bit256K65 tok/s0.3 sPDF: Qwen3 VL 8B Instruct on L40S report
2× RTX 50908-bit256K125 tok/s0.1 sPDF: Qwen3 VL 8B Instruct on 2× RTX 5090 report
4× RTX 30908-bit256K70 tok/s0.1 sPDF: Qwen3 VL 8B Instruct on 4× RTX 3090 report
RTX PRO 6000 Blackwell8-bit256K125 tok/s0.1 sPDF: Qwen3 VL 8B Instruct on RTX PRO 6000 Blackwell report
Mac mini M4 16 GB6-bit16K14 tok/s11 sPDF: Qwen3 VL 8B Instruct on Mac mini M4 16 GB report
Mac mini M4 24 GB8-bit40K11 tok/s11 sPDF: Qwen3 VL 8B Instruct on Mac mini M4 24 GB report
Mac mini M4 32 GB8-bit72K11 tok/s11 sPDF: Qwen3 VL 8B Instruct on Mac mini M4 32 GB report
Mac mini M4 Pro 24 GB8-bit40K24 tok/s5.4 sPDF: Qwen3 VL 8B Instruct on Mac mini M4 Pro 24 GB report
Mac mini M4 Pro 48 GB8-bit176K24 tok/s5.4 sPDF: Qwen3 VL 8B Instruct on Mac mini M4 Pro 48 GB report
Mac mini M4 Pro 64 GB8-bit256K24 tok/s5.4 sPDF: Qwen3 VL 8B Instruct on Mac mini M4 Pro 64 GB report
M4 Max 128 GB8-bit256K42 tok/s2.7 sPDF: Qwen3 VL 8B Instruct on M4 Max 128 GB report
M5 Max 128 GB8-bit256K46 tok/s0.8 sPDF: Qwen3 VL 8B Instruct on M5 Max 128 GB report
M3 Ultra 96 GB8-bit256K52 tok/s1.5 sPDF: Qwen3 VL 8B Instruct on M3 Ultra 96 GB report
M3 Ultra 256 GB8-bit256K52 tok/s1.5 sPDF: Qwen3 VL 8B Instruct on M3 Ultra 256 GB report
M3 Ultra 512 GB8-bit256K52 tok/s1.5 sPDF: Qwen3 VL 8B Instruct on M3 Ultra 512 GB report
M5 Ultra 512 GB8-bit256K72 tok/s0.5 sPDF: Qwen3 VL 8B Instruct on M5 Ultra 512 GB report
Ryzen AI Max+ 395 64 GB8-bit256K17 tok/s1.5 sPDF: Qwen3 VL 8B Instruct on Ryzen AI Max+ 395 64 GB report
Ryzen AI Max+ 395 128 GB8-bit256K17 tok/s1.5 sPDF: Qwen3 VL 8B Instruct on Ryzen AI Max+ 395 128 GB report
RTX Spark8-bit256K24 tok/s0.5 sPDF: Qwen3 VL 8B Instruct on RTX Spark report
DGX Spark8-bit256K22 tok/s0.5 sPDF: Qwen3 VL 8B Instruct on DGX Spark report
Dual DGX Spark8-bit256K22 tok/s0.3 sPDF: Qwen3 VL 8B Instruct on Dual DGX Spark report
DGX Station GB3008-bit256K354 tok/s0.1 sPDF: Qwen3 VL 8B Instruct on DGX Station GB300 report
2× RTX PRO 6000 Blackwell8-bit256K125 tok/s0.1 sPDF: Qwen3 VL 8B Instruct on 2× RTX PRO 6000 Blackwell report
4× RTX PRO 6000 Blackwell8-bit256K125 tok/s0.1 sPDF: Qwen3 VL 8B Instruct on 4× RTX PRO 6000 Blackwell report
8× RTX PRO 6000 Blackwell8-bit256K125 tok/s0.1 sPDF: Qwen3 VL 8B Instruct on 8× RTX PRO 6000 Blackwell report
A100 80 GB SXM8-bit256K140 tok/s0.3 sPDF: Qwen3 VL 8B Instruct on A100 80 GB SXM report
H100 SXM8-bit256K210 tok/s0.1 sPDF: Qwen3 VL 8B Instruct on H100 SXM report
4× H200 NVL8-bit256K274 tok/s0.1 sPDF: Qwen3 VL 8B Instruct on 4× H200 NVL report
8× H200 NVL8-bit256K274 tok/s0.1 sPDF: Qwen3 VL 8B Instruct 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-VL-8B-Instruct-GGUF
vLLM
vllm serve Qwen/Qwen3-VL-8B-Instruct

Questions

How much VRAM does Qwen3 VL 8B Instruct need?

Qwen3 VL 8B Instruct needs 5.0 GB for its weights at 4-bit, 6.9 GB with an 8K context and 10.3 GB with 32K. At 8-bit, 10.7 GB.

What is the minimum VRAM for Qwen3 VL 8B Instruct?

About 6.0 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 4060.

How big is Qwen3 VL 8B Instruct?

8.8B parameters: 16.3 GB of weights at 16-bit, 5.0 GB at 4-bit.

Can Qwen3 VL 8B Instruct run on an RTX 4090?

Yes. At 8-bit it holds 88K of context and writes about 75 tokens a second.

More models