Skip to content

Qwen 7B Chat VRAM requirements

How much VRAM Qwen 7B Chat needs
4-bit weights: 4.4 GBWith 8K context: 9.2 GBWith 32K context: 21.2 GB8-bit, 8K context: 12.5 GB
Parameters
7.7B
Longest context
32K
Released
Aug 2023
License
Tongyi Qianwen License Agreement
Strongest at
Not rated

Its licence (Tongyi Qianwen License Agreement) is not a standard open-source one: check the terms before commercial use.

How much VRAM Qwen 7B Chat needs

Memory Qwen 7B Chat needs at each precision and context length
PrecisionWeightsWith 8KWith 32K
16-bit14.4 GB19.2 GB31.2 GB
8-bit7.6 GB12.5 GB24.5 GB
6-bit5.9 GB10.8 GB22.8 GB
5-bit5.1 GB10.0 GB22.0 GB
4-bit4.4 GB9.2 GB21.2 GB
3-bit3.5 GB8.4 GB20.4 GB
2-bit2.7 GB7.6 GB19.6 GB
1.6-bit1.5 GB6.3 GB18.3 GB

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

GPUs and Macs that run Qwen 7B Chat

The best way Qwen 7B Chat runs on common GPUs, Macs and AI boxes
MachineBest fitContextSpeedFirst replyReport
RTX 4060Too bigPDF: Qwen 7B Chat on RTX 4060 report
RTX 5060Too bigPDF: Qwen 7B Chat on RTX 5060 report
RTX 3070Too bigPDF: Qwen 7B Chat on RTX 3070 report
RTX 3080 10 GBToo bigPDF: Qwen 7B Chat on RTX 3080 10 GB report
RTX 3060 12 GBToo bigPDF: Qwen 7B Chat on RTX 3060 12 GB report
Arc B580Too bigPDF: Qwen 7B Chat on Arc B580 report
RTX 4070Too bigPDF: Qwen 7B Chat on RTX 4070 report
RTX 4070 SuperToo bigPDF: Qwen 7B Chat on RTX 4070 Super report
RTX 5070Too bigPDF: Qwen 7B Chat on RTX 5070 report
RTX 4060 Ti 16 GB6-bit16K34 tok/s0.7 sPDF: Qwen 7B Chat on RTX 4060 Ti 16 GB report
RTX 5060 Ti 16 GB6-bit16K51 tok/s0.6 sPDF: Qwen 7B Chat on RTX 5060 Ti 16 GB report
Arc A770 16 GB6-bit16K44 tok/s0.7 sPDF: Qwen 7B Chat on Arc A770 16 GB report
Radeon RX 9070 XT6-bit16K60 tok/s0.4 sPDF: Qwen 7B Chat on Radeon RX 9070 XT report
RTX 4070 Ti Super6-bit16K75 tok/s0.3 sPDF: Qwen 7B Chat on RTX 4070 Ti Super report
RTX 4080 Super6-bit16K82 tok/s0.3 sPDF: Qwen 7B Chat on RTX 4080 Super report
RTX 5070 Ti6-bit16K98 tok/s0.3 sPDF: Qwen 7B Chat on RTX 5070 Ti report
RTX 50806-bit16K104 tok/s0.3 sPDF: Qwen 7B Chat on RTX 5080 report
Radeon RX 7900 XTX8-bit28K68 tok/s0.6 sPDF: Qwen 7B Chat on Radeon RX 7900 XTX report
RTX 30908-bit28K80 tok/s0.4 sPDF: Qwen 7B Chat on RTX 3090 report
RTX 40908-bit28K86 tok/s0.2 sPDF: Qwen 7B Chat on RTX 4090 report
Radeon AI PRO R97008-bit32K47 tok/s0.4 sPDF: Qwen 7B Chat on Radeon AI PRO R9700 report
RTX 50908-bit32K143 tok/s0.1 sPDF: Qwen 7B Chat on RTX 5090 report
2× RTX 30908-bit32K80 tok/s0.2 sPDF: Qwen 7B Chat on 2× RTX 3090 report
2× RTX 40908-bit32K86 tok/s0.1 sPDF: Qwen 7B Chat on 2× RTX 4090 report
RTX PRO 5000 Blackwell 48 GB8-bit32K111 tok/s0.2 sPDF: Qwen 7B Chat on RTX PRO 5000 Blackwell 48 GB report
RTX 6000 Ada8-bit32K82 tok/s0.2 sPDF: Qwen 7B Chat on RTX 6000 Ada report
L40S8-bit32K74 tok/s0.3 sPDF: Qwen 7B Chat on L40S report
2× RTX 50908-bit32K143 tok/s0.1 sPDF: Qwen 7B Chat on 2× RTX 5090 report
4× RTX 30908-bit32K80 tok/s0.1 sPDF: Qwen 7B Chat on 4× RTX 3090 report
RTX PRO 6000 Blackwell8-bit32K143 tok/s0.1 sPDF: Qwen 7B Chat on RTX PRO 6000 Blackwell report
Mac mini M4 16 GBToo bigPDF: Qwen 7B Chat on Mac mini M4 16 GB report
Mac mini M4 24 GB6-bit16K16 tok/s9.2 sPDF: Qwen 7B Chat on Mac mini M4 24 GB report
Mac mini M4 32 GB8-bit20K13 tok/s9.2 sPDF: Qwen 7B Chat on Mac mini M4 32 GB report
Mac mini M4 Pro 24 GB6-bit16K35 tok/s4.6 sPDF: Qwen 7B Chat on Mac mini M4 Pro 24 GB report
Mac mini M4 Pro 48 GB8-bit32K28 tok/s4.6 sPDF: Qwen 7B Chat on Mac mini M4 Pro 48 GB report
Mac mini M4 Pro 64 GB8-bit32K28 tok/s4.6 sPDF: Qwen 7B Chat on Mac mini M4 Pro 64 GB report
M4 Max 128 GB8-bit32K48 tok/s2.3 sPDF: Qwen 7B Chat on M4 Max 128 GB report
M5 Max 128 GB8-bit32K53 tok/s0.7 sPDF: Qwen 7B Chat on M5 Max 128 GB report
M3 Ultra 96 GB8-bit32K60 tok/s1.3 sPDF: Qwen 7B Chat on M3 Ultra 96 GB report
M3 Ultra 256 GB8-bit32K60 tok/s1.3 sPDF: Qwen 7B Chat on M3 Ultra 256 GB report
M3 Ultra 512 GB8-bit32K60 tok/s1.3 sPDF: Qwen 7B Chat on M3 Ultra 512 GB report
M5 Ultra 512 GB8-bit32K82 tok/s0.4 sPDF: Qwen 7B Chat on M5 Ultra 512 GB report
Ryzen AI Max+ 395 64 GB8-bit32K20 tok/s1.3 sPDF: Qwen 7B Chat on Ryzen AI Max+ 395 64 GB report
Ryzen AI Max+ 395 128 GB8-bit32K20 tok/s1.3 sPDF: Qwen 7B Chat on Ryzen AI Max+ 395 128 GB report
RTX Spark8-bit32K27 tok/s0.5 sPDF: Qwen 7B Chat on RTX Spark report
DGX Spark8-bit32K25 tok/s0.5 sPDF: Qwen 7B Chat on DGX Spark report
Dual DGX Spark8-bit32K25 tok/s0.3 sPDF: Qwen 7B Chat on Dual DGX Spark report
DGX Station GB3008-bit32K404 tok/s0.1 sPDF: Qwen 7B Chat on DGX Station GB300 report
2× RTX PRO 6000 Blackwell8-bit32K143 tok/s0.1 sPDF: Qwen 7B Chat on 2× RTX PRO 6000 Blackwell report
4× RTX PRO 6000 Blackwell8-bit32K143 tok/s0.1 sPDF: Qwen 7B Chat on 4× RTX PRO 6000 Blackwell report
8× RTX PRO 6000 Blackwell8-bit32K143 tok/s0.1 sPDF: Qwen 7B Chat on 8× RTX PRO 6000 Blackwell report
A100 80 GB SXM8-bit32K160 tok/s0.3 sPDF: Qwen 7B Chat on A100 80 GB SXM report
H100 SXM8-bit32K239 tok/s0.1 sPDF: Qwen 7B Chat on H100 SXM report
4× H200 NVL8-bit32K312 tok/s0.1 sPDF: Qwen 7B Chat on 4× H200 NVL report
8× H200 NVL8-bit32K312 tok/s0.1 sPDF: Qwen 7B Chat 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 mradermacher/Qwen-7B-Chat-GGUF
vLLM
vllm serve Qwen/Qwen-7B-Chat

Questions

How much VRAM does Qwen 7B Chat need?

Qwen 7B Chat needs 4.4 GB for its weights at 4-bit, 9.2 GB with an 8K context and 21.2 GB with 32K. At 8-bit, 12.5 GB.

What is the minimum VRAM for Qwen 7B Chat?

About 8.4 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 Ti 16 GB.

How big is Qwen 7B Chat?

7.7B parameters: 14.4 GB of weights at 16-bit, 4.4 GB at 4-bit.

Can Qwen 7B Chat run on an RTX 4090?

Yes. At 8-bit it holds 28K of context and writes about 86 tokens a second.

More models