Skip to content

WeVisOne VL 4B VRAM requirements

How much VRAM WeVisOne VL 4B needs
4-bit weights: 2.7 GBWith 8K context: 4.7 GBWith 32K context: 8.0 GB8-bit, 8K context: 6.7 GB
Parameters
4.8B
Longest context
256K
Released
Oct 2026
License
Apache 2.0
Strongest at
Text and images

How much VRAM WeVisOne VL 4B needs

Memory WeVisOne VL 4B needs at each precision and context length
PrecisionWeightsWith 8KWith 32KWith 128K
16-bit9.0 GB10.9 GB14.3 GB27.8 GB
8-bit4.8 GB6.7 GB10.1 GB23.6 GB
6-bit3.7 GB5.6 GB9.0 GB22.5 GB
5-bit3.2 GB5.1 GB8.5 GB22.0 GB
4-bit2.7 GB4.7 GB8.0 GB21.5 GB
3-bit2.2 GB4.1 GB7.5 GB21.0 GB
2-bit1.7 GB3.6 GB7.0 GB20.5 GB
1.6-bit950 MB2.9 GB6.2 GB19.7 GB

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

GPUs and Macs that run WeVisOne VL 4B

The best way WeVisOne VL 4B runs on common GPUs, Macs and AI boxes
MachineBest fitContextSpeedFirst replyReport
RTX 40606-bit20K50 tok/s0.6 sPDF: WeVisOne VL 4B on RTX 4060 report
RTX 50606-bit20K79 tok/s0.5 sPDF: WeVisOne VL 4B on RTX 5060 report
RTX 30706-bit20K79 tok/s0.5 sPDF: WeVisOne VL 4B on RTX 3070 report
RTX 3080 10 GB8-bit24K101 tok/s0.3 sPDF: WeVisOne VL 4B on RTX 3080 10 GB report
RTX 3060 12 GB8-bit40K51 tok/s0.7 sPDF: WeVisOne VL 4B on RTX 3060 12 GB report
Arc B5808-bit40K44 tok/s0.5 sPDF: WeVisOne VL 4B on Arc B580 report
RTX 40708-bit40K69 tok/s0.3 sPDF: WeVisOne VL 4B on RTX 4070 report
RTX 4070 Super8-bit40K69 tok/s0.3 sPDF: WeVisOne VL 4B on RTX 4070 Super report
RTX 50708-bit40K90 tok/s0.3 sPDF: WeVisOne VL 4B on RTX 5070 report
RTX 4060 Ti 16 GB8-bit64K41 tok/s0.4 sPDF: WeVisOne VL 4B on RTX 4060 Ti 16 GB report
RTX 5060 Ti 16 GB8-bit64K62 tok/s0.4 sPDF: WeVisOne VL 4B on RTX 5060 Ti 16 GB report
Arc A770 16 GB8-bit64K53 tok/s0.4 sPDF: WeVisOne VL 4B on Arc A770 16 GB report
Radeon RX 9070 XT8-bit64K71 tok/s0.3 sPDF: WeVisOne VL 4B on Radeon RX 9070 XT report
RTX 4070 Ti Super8-bit64K90 tok/s0.2 sPDF: WeVisOne VL 4B on RTX 4070 Ti Super report
RTX 4080 Super8-bit64K98 tok/s0.2 sPDF: WeVisOne VL 4B on RTX 4080 Super report
RTX 5070 Ti8-bit64K116 tok/s0.2 sPDF: WeVisOne VL 4B on RTX 5070 Ti report
RTX 50808-bit64K123 tok/s0.2 sPDF: WeVisOne VL 4B on RTX 5080 report
Radeon RX 7900 XTX8-bit120K100 tok/s0.4 sPDF: WeVisOne VL 4B on Radeon RX 7900 XTX report
RTX 30908-bit120K121 tok/s0.3 sPDF: WeVisOne VL 4B on RTX 3090 report
RTX 40908-bit120K129 tok/s0.1 sPDF: WeVisOne VL 4B on RTX 4090 report
Radeon AI PRO R97008-bit176K71 tok/s0.3 sPDF: WeVisOne VL 4B on Radeon AI PRO R9700 report
RTX 50908-bit176K207 tok/s0.1 sPDF: WeVisOne VL 4B on RTX 5090 report
2× RTX 30908-bit256K121 tok/s0.1 sPDF: WeVisOne VL 4B on 2× RTX 3090 report
2× RTX 40908-bit256K129 tok/s0.1 sPDF: WeVisOne VL 4B on 2× RTX 4090 report
RTX PRO 5000 Blackwell 48 GB8-bit256K164 tok/s0.1 sPDF: WeVisOne VL 4B on RTX PRO 5000 Blackwell 48 GB report
RTX 6000 Ada8-bit256K123 tok/s0.1 sPDF: WeVisOne VL 4B on RTX 6000 Ada report
L40S8-bit256K113 tok/s0.2 sPDF: WeVisOne VL 4B on L40S report
2× RTX 50908-bit256K207 tok/s0.1 sPDF: WeVisOne VL 4B on 2× RTX 5090 report
4× RTX 30908-bit256K121 tok/s0.1 sPDF: WeVisOne VL 4B on 4× RTX 3090 report
RTX PRO 6000 Blackwell8-bit256K207 tok/s0.1 sPDF: WeVisOne VL 4B on RTX PRO 6000 Blackwell report
Mac mini M4 16 GB8-bit32K20 tok/s5.8 sPDF: WeVisOne VL 4B on Mac mini M4 16 GB report
Mac mini M4 24 GB8-bit64K20 tok/s5.8 sPDF: WeVisOne VL 4B on Mac mini M4 24 GB report
Mac mini M4 32 GB8-bit104K20 tok/s5.8 sPDF: WeVisOne VL 4B on Mac mini M4 32 GB report
Mac mini M4 Pro 24 GB8-bit64K42 tok/s2.9 sPDF: WeVisOne VL 4B on Mac mini M4 Pro 24 GB report
Mac mini M4 Pro 48 GB8-bit200K42 tok/s2.9 sPDF: WeVisOne VL 4B on Mac mini M4 Pro 48 GB report
Mac mini M4 Pro 64 GB8-bit256K42 tok/s2.9 sPDF: WeVisOne VL 4B on Mac mini M4 Pro 64 GB report
M4 Max 128 GB8-bit256K69 tok/s1.5 sPDF: WeVisOne VL 4B on M4 Max 128 GB report
M5 Max 128 GB8-bit256K76 tok/s0.4 sPDF: WeVisOne VL 4B on M5 Max 128 GB report
M3 Ultra 96 GB8-bit256K85 tok/s0.8 sPDF: WeVisOne VL 4B on M3 Ultra 96 GB report
M3 Ultra 256 GB8-bit256K85 tok/s0.8 sPDF: WeVisOne VL 4B on M3 Ultra 256 GB report
M3 Ultra 512 GB8-bit256K85 tok/s0.8 sPDF: WeVisOne VL 4B on M3 Ultra 512 GB report
M5 Ultra 512 GB8-bit256K112 tok/s0.3 sPDF: WeVisOne VL 4B on M5 Ultra 512 GB report
Ryzen AI Max+ 395 64 GB8-bit256K31 tok/s0.8 sPDF: WeVisOne VL 4B on Ryzen AI Max+ 395 64 GB report
Ryzen AI Max+ 395 128 GB8-bit256K31 tok/s0.8 sPDF: WeVisOne VL 4B on Ryzen AI Max+ 395 128 GB report
RTX Spark8-bit256K42 tok/s0.3 sPDF: WeVisOne VL 4B on RTX Spark report
DGX Spark8-bit256K39 tok/s0.3 sPDF: WeVisOne VL 4B on DGX Spark report
Dual DGX Spark8-bit256K39 tok/s0.2 sPDF: WeVisOne VL 4B on Dual DGX Spark report
DGX Station GB3008-bit256K493 tok/s0.1 sPDF: WeVisOne VL 4B on DGX Station GB300 report
2× RTX PRO 6000 Blackwell8-bit256K207 tok/s0.1 sPDF: WeVisOne VL 4B on 2× RTX PRO 6000 Blackwell report
4× RTX PRO 6000 Blackwell8-bit256K207 tok/s0.1 sPDF: WeVisOne VL 4B on 4× RTX PRO 6000 Blackwell report
8× RTX PRO 6000 Blackwell8-bit256K207 tok/s0.1 sPDF: WeVisOne VL 4B on 8× RTX PRO 6000 Blackwell report
A100 80 GB SXM8-bit256K228 tok/s0.2 sPDF: WeVisOne VL 4B on A100 80 GB SXM report
H100 SXM8-bit256K323 tok/s0.1 sPDF: WeVisOne VL 4B on H100 SXM report
4× H200 NVL8-bit256K403 tok/s0.1 sPDF: WeVisOne VL 4B on 4× H200 NVL report
8× H200 NVL8-bit256K403 tok/s0.1 sPDF: WeVisOne VL 4B 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/WeVisOne-VL-4B-GGUF
vLLM
vllm serve tencent/WeVisOne-VL-4B

Questions

How much VRAM does WeVisOne VL 4B need?

WeVisOne VL 4B needs 2.7 GB for its weights at 4-bit, 4.7 GB with an 8K context and 8.0 GB with 32K. At 8-bit, 6.7 GB.

What is the minimum VRAM for WeVisOne VL 4B?

About 4.1 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 WeVisOne VL 4B?

4.8B parameters: 9.0 GB of weights at 16-bit, 2.7 GB at 4-bit.

Can WeVisOne VL 4B run on an RTX 4090?

Yes. At 8-bit it holds 120K of context and writes about 129 tokens a second.

More models