Skip to content

Qwen3Guard Gen 0.6B VRAM requirements

How much VRAM Qwen3Guard Gen 0.6B needs
4-bit weights: 435 MBWith 8K context: 2.1 GBWith 32K context: 4.7 GB8-bit, 8K context: 2.4 GB
Parameters
752M
Longest context
32K
Released
Sep 2025
License
Apache 2.0
Strongest at
Not rated

How much VRAM Qwen3Guard Gen 0.6B needs

Memory Qwen3Guard Gen 0.6B needs at each precision and context length
PrecisionWeightsWith 8KWith 32K
16-bit1.4 GB3.0 GB5.7 GB
8-bit762 MB2.4 GB5.0 GB
6-bit588 MB2.2 GB4.8 GB
5-bit510 MB2.1 GB4.8 GB
4-bit435 MB2.1 GB4.7 GB
3-bit350 MB2.0 GB4.6 GB
2-bit269 MB1.9 GB4.5 GB
1.6-bit148 MB1.8 GB4.4 GB

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

GPUs and Macs that run Qwen3Guard Gen 0.6B

The best way Qwen3Guard Gen 0.6B runs on common GPUs, Macs and AI boxes
MachineBest fitContextSpeedFirst replyReport
RTX 40608-bit32K199 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on RTX 4060 report
RTX 50608-bit32K295 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on RTX 5060 report
RTX 30708-bit32K295 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on RTX 3070 report
RTX 3080 10 GB8-bit32K427 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on RTX 3080 10 GB report
RTX 3060 12 GB8-bit32K249 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on RTX 3060 12 GB report
Arc B5808-bit32K193 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on Arc B580 report
RTX 40708-bit32K322 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on RTX 4070 report
RTX 4070 Super8-bit32K322 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on RTX 4070 Super report
RTX 50708-bit32K394 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on RTX 5070 report
RTX 4060 Ti 16 GB8-bit32K208 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on RTX 4060 Ti 16 GB report
RTX 5060 Ti 16 GB8-bit32K295 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on RTX 5060 Ti 16 GB report
Arc A770 16 GB8-bit32K221 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on Arc A770 16 GB report
Radeon RX 9070 XT8-bit32K287 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on Radeon RX 9070 XT report
RTX 4070 Ti Super8-bit32K394 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on RTX 4070 Ti Super report
RTX 4080 Super8-bit32K418 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on RTX 4080 Super report
RTX 5070 Ti8-bit32K473 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on RTX 5070 Ti report
RTX 50808-bit32K493 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on RTX 5080 report
Radeon RX 7900 XTX8-bit32K359 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on Radeon RX 7900 XTX report
RTX 30908-bit32K486 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on RTX 3090 report
RTX 40908-bit32K507 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on RTX 4090 report
Radeon AI PRO R97008-bit32K287 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on Radeon AI PRO R9700 report
RTX 50908-bit32K677 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on RTX 5090 report
2× RTX 30908-bit32K486 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on 2× RTX 3090 report
2× RTX 40908-bit32K507 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on 2× RTX 4090 report
RTX PRO 5000 Blackwell 48 GB8-bit32K592 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on RTX PRO 5000 Blackwell 48 GB report
RTX 6000 Ada8-bit32K493 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on RTX 6000 Ada report
L40S8-bit32K463 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on L40S report
2× RTX 50908-bit32K677 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on 2× RTX 5090 report
4× RTX 30908-bit32K486 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on 4× RTX 3090 report
RTX PRO 6000 Blackwell8-bit32K677 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on RTX PRO 6000 Blackwell report
Mac mini M4 16 GB8-bit32K97 tok/s0.9 sPDF: Qwen3Guard Gen 0.6B on Mac mini M4 16 GB report
Mac mini M4 24 GB8-bit32K97 tok/s0.9 sPDF: Qwen3Guard Gen 0.6B on Mac mini M4 24 GB report
Mac mini M4 32 GB8-bit32K97 tok/s0.9 sPDF: Qwen3Guard Gen 0.6B on Mac mini M4 32 GB report
Mac mini M4 Pro 24 GB8-bit32K172 tok/s0.4 sPDF: Qwen3Guard Gen 0.6B on Mac mini M4 Pro 24 GB report
Mac mini M4 Pro 48 GB8-bit32K172 tok/s0.4 sPDF: Qwen3Guard Gen 0.6B on Mac mini M4 Pro 48 GB report
Mac mini M4 Pro 64 GB8-bit32K172 tok/s0.4 sPDF: Qwen3Guard Gen 0.6B on Mac mini M4 Pro 64 GB report
M4 Max 128 GB8-bit32K239 tok/s0.2 sPDF: Qwen3Guard Gen 0.6B on M4 Max 128 GB report
M5 Max 128 GB8-bit32K252 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on M5 Max 128 GB report
M3 Ultra 96 GB8-bit32K267 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on M3 Ultra 96 GB report
M3 Ultra 256 GB8-bit32K267 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on M3 Ultra 256 GB report
M3 Ultra 512 GB8-bit32K267 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on M3 Ultra 512 GB report
M5 Ultra 512 GB8-bit32K306 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on M5 Ultra 512 GB report
Ryzen AI Max+ 395 64 GB8-bit32K152 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on Ryzen AI Max+ 395 64 GB report
Ryzen AI Max+ 395 128 GB8-bit32K152 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on Ryzen AI Max+ 395 128 GB report
RTX Spark8-bit32K215 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on RTX Spark report
DGX Spark8-bit32K199 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on DGX Spark report
Dual DGX Spark8-bit32K199 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on Dual DGX Spark report
DGX Station GB3008-bit32K999 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on DGX Station GB300 report
2× RTX PRO 6000 Blackwell8-bit32K677 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on 2× RTX PRO 6000 Blackwell report
4× RTX PRO 6000 Blackwell8-bit32K677 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on 4× RTX PRO 6000 Blackwell report
8× RTX PRO 6000 Blackwell8-bit32K677 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on 8× RTX PRO 6000 Blackwell report
A100 80 GB SXM8-bit32K715 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on A100 80 GB SXM report
H100 SXM8-bit32K847 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on H100 SXM report
4× H200 NVL8-bit32K928 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B on 4× H200 NVL report
8× H200 NVL8-bit32K928 tok/s0.1 sPDF: Qwen3Guard Gen 0.6B 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/Qwen3Guard-Gen-0.6B-GGUF
vLLM
vllm serve Qwen/Qwen3Guard-Gen-0.6B

Questions

How much VRAM does Qwen3Guard Gen 0.6B need?

Qwen3Guard Gen 0.6B needs 435 MB for its weights at 4-bit, 2.1 GB with an 8K context and 4.7 GB with 32K. At 8-bit, 2.4 GB.

What is the minimum VRAM for Qwen3Guard Gen 0.6B?

About 2.0 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 Qwen3Guard Gen 0.6B?

752M parameters: 1.4 GB of weights at 16-bit, 435 MB at 4-bit.

Can Qwen3Guard Gen 0.6B run on an RTX 4090?

Yes. At 8-bit it holds 32K of context and writes about 507 tokens a second.

More models