Skip to content

Granite Vision 4.1 4B VRAM requirements

How much VRAM Granite Vision 4.1 4B needs
4-bit weights: 2.3 GBWith 8K context: 3.6 GBWith 32K context: 5.5 GB8-bit, 8K context: 5.3 GB
Parameters
4B
Longest context
128K
Released
Apr 2026
License
Apache 2.0
Strongest at
Text and images

How much VRAM Granite Vision 4.1 4B needs

Memory Granite Vision 4.1 4B needs at each precision and context length
PrecisionWeightsWith 8KWith 32KWith 128K
16-bit7.4 GB8.8 GB10.7 GB18.2 GB
8-bit4.0 GB5.3 GB7.2 GB14.7 GB
6-bit3.1 GB4.4 GB6.3 GB13.8 GB
5-bit2.6 GB4.0 GB5.9 GB13.4 GB
4-bit2.3 GB3.6 GB5.5 GB13.0 GB
3-bit1.8 GB3.2 GB5.0 GB12.5 GB
2-bit1.4 GB2.7 GB4.6 GB12.1 GB
1.6-bit786 MB2.1 GB4.0 GB11.5 GB

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

GPUs and Macs that run Granite Vision 4.1 4B

The best way Granite Vision 4.1 4B runs on common GPUs, Macs and AI boxes
MachineBest fitContextSpeedFirst replyReport
RTX 40608-bit36K43 tok/s0.6 sPDF: Granite Vision 4.1 4B on RTX 4060 report
RTX 50608-bit36K68 tok/s0.4 sPDF: Granite Vision 4.1 4B on RTX 5060 report
RTX 30708-bit36K68 tok/s0.4 sPDF: Granite Vision 4.1 4B on RTX 3070 report
RTX 3080 10 GB8-bit60K109 tok/s0.3 sPDF: Granite Vision 4.1 4B on RTX 3080 10 GB report
RTX 3060 12 GB8-bit80K55 tok/s0.7 sPDF: Granite Vision 4.1 4B on RTX 3060 12 GB report
Arc B5808-bit80K48 tok/s0.5 sPDF: Granite Vision 4.1 4B on Arc B580 report
RTX 40708-bit80K76 tok/s0.3 sPDF: Granite Vision 4.1 4B on RTX 4070 report
RTX 4070 Super8-bit80K76 tok/s0.2 sPDF: Granite Vision 4.1 4B on RTX 4070 Super report
RTX 50708-bit80K98 tok/s0.3 sPDF: Granite Vision 4.1 4B on RTX 5070 report
RTX 4060 Ti 16 GB8-bit128K45 tok/s0.4 sPDF: Granite Vision 4.1 4B on RTX 4060 Ti 16 GB report
RTX 5060 Ti 16 GB8-bit128K68 tok/s0.4 sPDF: Granite Vision 4.1 4B on RTX 5060 Ti 16 GB report
Arc A770 16 GB8-bit128K57 tok/s0.4 sPDF: Granite Vision 4.1 4B on Arc A770 16 GB report
Radeon RX 9070 XT8-bit128K77 tok/s0.2 sPDF: Granite Vision 4.1 4B on Radeon RX 9070 XT report
RTX 4070 Ti Super8-bit128K98 tok/s0.2 sPDF: Granite Vision 4.1 4B on RTX 4070 Ti Super report
RTX 4080 Super8-bit128K106 tok/s0.2 sPDF: Granite Vision 4.1 4B on RTX 4080 Super report
RTX 5070 Ti8-bit128K125 tok/s0.2 sPDF: Granite Vision 4.1 4B on RTX 5070 Ti report
RTX 50808-bit128K133 tok/s0.2 sPDF: Granite Vision 4.1 4B on RTX 5080 report
Radeon RX 7900 XTX8-bit128K107 tok/s0.4 sPDF: Granite Vision 4.1 4B on Radeon RX 7900 XTX report
RTX 30908-bit128K130 tok/s0.2 sPDF: Granite Vision 4.1 4B on RTX 3090 report
RTX 40908-bit128K139 tok/s0.1 sPDF: Granite Vision 4.1 4B on RTX 4090 report
Radeon AI PRO R97008-bit128K77 tok/s0.2 sPDF: Granite Vision 4.1 4B on Radeon AI PRO R9700 report
RTX 50908-bit128K218 tok/s0.1 sPDF: Granite Vision 4.1 4B on RTX 5090 report
2× RTX 30908-bit128K130 tok/s0.1 sPDF: Granite Vision 4.1 4B on 2× RTX 3090 report
2× RTX 40908-bit128K139 tok/s0.1 sPDF: Granite Vision 4.1 4B on 2× RTX 4090 report
RTX PRO 5000 Blackwell 48 GB8-bit128K175 tok/s0.1 sPDF: Granite Vision 4.1 4B on RTX PRO 5000 Blackwell 48 GB report
RTX 6000 Ada8-bit128K133 tok/s0.1 sPDF: Granite Vision 4.1 4B on RTX 6000 Ada report
L40S8-bit128K122 tok/s0.1 sPDF: Granite Vision 4.1 4B on L40S report
2× RTX 50908-bit128K218 tok/s0.1 sPDF: Granite Vision 4.1 4B on 2× RTX 5090 report
4× RTX 30908-bit128K130 tok/s0.1 sPDF: Granite Vision 4.1 4B on 4× RTX 3090 report
RTX PRO 6000 Blackwell8-bit128K218 tok/s0.1 sPDF: Granite Vision 4.1 4B on RTX PRO 6000 Blackwell report
Mac mini M4 16 GB8-bit72K22 tok/s5.4 sPDF: Granite Vision 4.1 4B on Mac mini M4 16 GB report
Mac mini M4 24 GB8-bit128K22 tok/s5.4 sPDF: Granite Vision 4.1 4B on Mac mini M4 24 GB report
Mac mini M4 32 GB8-bit128K22 tok/s5.4 sPDF: Granite Vision 4.1 4B on Mac mini M4 32 GB report
Mac mini M4 Pro 24 GB8-bit128K45 tok/s2.7 sPDF: Granite Vision 4.1 4B on Mac mini M4 Pro 24 GB report
Mac mini M4 Pro 48 GB8-bit128K45 tok/s2.7 sPDF: Granite Vision 4.1 4B on Mac mini M4 Pro 48 GB report
Mac mini M4 Pro 64 GB8-bit128K45 tok/s2.7 sPDF: Granite Vision 4.1 4B on Mac mini M4 Pro 64 GB report
M4 Max 128 GB8-bit128K74 tok/s1.4 sPDF: Granite Vision 4.1 4B on M4 Max 128 GB report
M5 Max 128 GB8-bit128K80 tok/s0.4 sPDF: Granite Vision 4.1 4B on M5 Max 128 GB report
M3 Ultra 96 GB8-bit128K89 tok/s0.8 sPDF: Granite Vision 4.1 4B on M3 Ultra 96 GB report
M3 Ultra 256 GB8-bit128K89 tok/s0.8 sPDF: Granite Vision 4.1 4B on M3 Ultra 256 GB report
M3 Ultra 512 GB8-bit128K89 tok/s0.8 sPDF: Granite Vision 4.1 4B on M3 Ultra 512 GB report
M5 Ultra 512 GB8-bit128K116 tok/s0.2 sPDF: Granite Vision 4.1 4B on M5 Ultra 512 GB report
Ryzen AI Max+ 395 64 GB8-bit128K34 tok/s0.8 sPDF: Granite Vision 4.1 4B on Ryzen AI Max+ 395 64 GB report
Ryzen AI Max+ 395 128 GB8-bit128K34 tok/s0.8 sPDF: Granite Vision 4.1 4B on Ryzen AI Max+ 395 128 GB report
RTX Spark8-bit128K47 tok/s0.3 sPDF: Granite Vision 4.1 4B on RTX Spark report
DGX Spark8-bit128K43 tok/s0.3 sPDF: Granite Vision 4.1 4B on DGX Spark report
Dual DGX Spark8-bit128K43 tok/s0.2 sPDF: Granite Vision 4.1 4B on Dual DGX Spark report
DGX Station GB3008-bit128K487 tok/s0.1 sPDF: Granite Vision 4.1 4B on DGX Station GB300 report
2× RTX PRO 6000 Blackwell8-bit128K218 tok/s0.1 sPDF: Granite Vision 4.1 4B on 2× RTX PRO 6000 Blackwell report
4× RTX PRO 6000 Blackwell8-bit128K218 tok/s0.1 sPDF: Granite Vision 4.1 4B on 4× RTX PRO 6000 Blackwell report
8× RTX PRO 6000 Blackwell8-bit128K218 tok/s0.1 sPDF: Granite Vision 4.1 4B on 8× RTX PRO 6000 Blackwell report
A100 80 GB SXM8-bit128K239 tok/s0.2 sPDF: Granite Vision 4.1 4B on A100 80 GB SXM report
H100 SXM8-bit128K332 tok/s0.1 sPDF: Granite Vision 4.1 4B on H100 SXM report
4× H200 NVL8-bit128K406 tok/s0.1 sPDF: Granite Vision 4.1 4B on 4× H200 NVL report
8× H200 NVL8-bit128K406 tok/s0.1 sPDF: Granite Vision 4.1 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 ibm-granite/granite-vision-4.1-4b-GGUF
vLLM
vllm serve ibm-granite/granite-vision-4.1-4b

Questions

How much VRAM does Granite Vision 4.1 4B need?

Granite Vision 4.1 4B needs 2.3 GB for its weights at 4-bit, 3.6 GB with an 8K context and 5.5 GB with 32K. At 8-bit, 5.3 GB.

What is the minimum VRAM for Granite Vision 4.1 4B?

About 3.2 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 Granite Vision 4.1 4B?

4B parameters: 7.4 GB of weights at 16-bit, 2.3 GB at 4-bit.

Can Granite Vision 4.1 4B run on an RTX 4090?

Yes. At 8-bit it holds 128K of context and writes about 139 tokens a second.

More models