Skip to content

InternVL3 8B VRAM requirements

How much VRAM InternVL3 8B needs
4-bit weights: 4.5 GBWith 8K context: 5.8 GBWith 32K context: 7.1 GB8-bit, 8K context: 9.1 GB
Parameters
7.9B
Longest context
32K
Released
Apr 2025
License
Apache 2.0
Strongest at
Text and images

How much VRAM InternVL3 8B needs

Memory InternVL3 8B needs at each precision and context length
PrecisionWeightsWith 8KWith 32K
16-bit14.8 GB16.1 GB17.4 GB
8-bit7.9 GB9.1 GB10.5 GB
6-bit6.1 GB7.4 GB8.7 GB
5-bit5.3 GB6.5 GB7.9 GB
4-bit4.5 GB5.8 GB7.1 GB
3-bit3.6 GB4.9 GB6.2 GB
2-bit2.8 GB4.1 GB5.4 GB
1.6-bit1.5 GB2.8 GB4.1 GB

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

GPUs and Macs that run InternVL3 8B

The best way InternVL3 8B runs on common GPUs, Macs and AI boxes
MachineBest fitContextSpeedFirst replyReport
RTX 40605-bit24K35 tok/s1.0 sPDF: InternVL3 8B on RTX 4060 report
RTX 50605-bit24K57 tok/s0.8 sPDF: InternVL3 8B on RTX 5060 report
RTX 30705-bit24K57 tok/s0.8 sPDF: InternVL3 8B on RTX 3070 report
RTX 3080 10 GB6-bit32K82 tok/s0.5 sPDF: InternVL3 8B on RTX 3080 10 GB report
RTX 3060 12 GB8-bit32K31 tok/s1.2 sPDF: InternVL3 8B on RTX 3060 12 GB report
Arc B5808-bit32K28 tok/s0.8 sPDF: InternVL3 8B on Arc B580 report
RTX 40708-bit32K43 tok/s0.5 sPDF: InternVL3 8B on RTX 4070 report
RTX 4070 Super8-bit32K43 tok/s0.4 sPDF: InternVL3 8B on RTX 4070 Super report
RTX 50708-bit32K57 tok/s0.5 sPDF: InternVL3 8B on RTX 5070 report
RTX 4060 Ti 16 GB8-bit32K25 tok/s0.7 sPDF: InternVL3 8B on RTX 4060 Ti 16 GB report
RTX 5060 Ti 16 GB8-bit32K39 tok/s0.6 sPDF: InternVL3 8B on RTX 5060 Ti 16 GB report
Arc A770 16 GB8-bit32K34 tok/s0.7 sPDF: InternVL3 8B on Arc A770 16 GB report
Radeon RX 9070 XT8-bit32K46 tok/s0.4 sPDF: InternVL3 8B on Radeon RX 9070 XT report
RTX 4070 Ti Super8-bit32K57 tok/s0.3 sPDF: InternVL3 8B on RTX 4070 Ti Super report
RTX 4080 Super8-bit32K62 tok/s0.3 sPDF: InternVL3 8B on RTX 4080 Super report
RTX 5070 Ti8-bit32K75 tok/s0.3 sPDF: InternVL3 8B on RTX 5070 Ti report
RTX 50808-bit32K80 tok/s0.3 sPDF: InternVL3 8B on RTX 5080 report
Radeon RX 7900 XTX8-bit32K66 tok/s0.7 sPDF: InternVL3 8B on Radeon RX 7900 XTX report
RTX 30908-bit32K78 tok/s0.4 sPDF: InternVL3 8B on RTX 3090 report
RTX 40908-bit32K83 tok/s0.2 sPDF: InternVL3 8B on RTX 4090 report
Radeon AI PRO R97008-bit32K46 tok/s0.4 sPDF: InternVL3 8B on Radeon AI PRO R9700 report
RTX 50908-bit32K141 tok/s0.1 sPDF: InternVL3 8B on RTX 5090 report
2× RTX 30908-bit32K78 tok/s0.2 sPDF: InternVL3 8B on 2× RTX 3090 report
2× RTX 40908-bit32K83 tok/s0.1 sPDF: InternVL3 8B on 2× RTX 4090 report
RTX PRO 5000 Blackwell 48 GB8-bit32K109 tok/s0.2 sPDF: InternVL3 8B on RTX PRO 5000 Blackwell 48 GB report
RTX 6000 Ada8-bit32K80 tok/s0.2 sPDF: InternVL3 8B on RTX 6000 Ada report
L40S8-bit32K72 tok/s0.3 sPDF: InternVL3 8B on L40S report
2× RTX 50908-bit32K141 tok/s0.1 sPDF: InternVL3 8B on 2× RTX 5090 report
4× RTX 30908-bit32K78 tok/s0.1 sPDF: InternVL3 8B on 4× RTX 3090 report
RTX PRO 6000 Blackwell8-bit32K141 tok/s0.1 sPDF: InternVL3 8B on RTX PRO 6000 Blackwell report
Mac mini M4 16 GB8-bit28K12 tok/s9.7 sPDF: InternVL3 8B on Mac mini M4 16 GB report
Mac mini M4 24 GB8-bit32K12 tok/s9.7 sPDF: InternVL3 8B on Mac mini M4 24 GB report
Mac mini M4 32 GB8-bit32K12 tok/s9.7 sPDF: InternVL3 8B on Mac mini M4 32 GB report
Mac mini M4 Pro 24 GB8-bit32K27 tok/s4.8 sPDF: InternVL3 8B on Mac mini M4 Pro 24 GB report
Mac mini M4 Pro 48 GB8-bit32K27 tok/s4.8 sPDF: InternVL3 8B on Mac mini M4 Pro 48 GB report
Mac mini M4 Pro 64 GB8-bit32K27 tok/s4.8 sPDF: InternVL3 8B on Mac mini M4 Pro 64 GB report
M4 Max 128 GB8-bit32K47 tok/s2.4 sPDF: InternVL3 8B on M4 Max 128 GB report
M5 Max 128 GB8-bit32K52 tok/s0.7 sPDF: InternVL3 8B on M5 Max 128 GB report
M3 Ultra 96 GB8-bit32K59 tok/s1.4 sPDF: InternVL3 8B on M3 Ultra 96 GB report
M3 Ultra 256 GB8-bit32K59 tok/s1.4 sPDF: InternVL3 8B on M3 Ultra 256 GB report
M3 Ultra 512 GB8-bit32K59 tok/s1.4 sPDF: InternVL3 8B on M3 Ultra 512 GB report
M5 Ultra 512 GB8-bit32K81 tok/s0.4 sPDF: InternVL3 8B on M5 Ultra 512 GB report
Ryzen AI Max+ 395 64 GB8-bit32K19 tok/s1.4 sPDF: InternVL3 8B on Ryzen AI Max+ 395 64 GB report
Ryzen AI Max+ 395 128 GB8-bit32K19 tok/s1.4 sPDF: InternVL3 8B on Ryzen AI Max+ 395 128 GB report
RTX Spark8-bit32K26 tok/s0.5 sPDF: InternVL3 8B on RTX Spark report
DGX Spark8-bit32K24 tok/s0.5 sPDF: InternVL3 8B on DGX Spark report
Dual DGX Spark8-bit32K24 tok/s0.3 sPDF: InternVL3 8B on Dual DGX Spark report
DGX Station GB3008-bit32K413 tok/s0.1 sPDF: InternVL3 8B on DGX Station GB300 report
2× RTX PRO 6000 Blackwell8-bit32K141 tok/s0.1 sPDF: InternVL3 8B on 2× RTX PRO 6000 Blackwell report
4× RTX PRO 6000 Blackwell8-bit32K141 tok/s0.1 sPDF: InternVL3 8B on 4× RTX PRO 6000 Blackwell report
8× RTX PRO 6000 Blackwell8-bit32K141 tok/s0.1 sPDF: InternVL3 8B on 8× RTX PRO 6000 Blackwell report
A100 80 GB SXM8-bit32K157 tok/s0.3 sPDF: InternVL3 8B on A100 80 GB SXM report
H100 SXM8-bit32K238 tok/s0.1 sPDF: InternVL3 8B on H100 SXM report
4× H200 NVL8-bit32K314 tok/s0.1 sPDF: InternVL3 8B on 4× H200 NVL report
8× H200 NVL8-bit32K314 tok/s0.1 sPDF: InternVL3 8B 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/InternVL3-8B-i1-GGUF
vLLM
vllm serve OpenGVLab/InternVL3-8B

Questions

How much VRAM does InternVL3 8B need?

InternVL3 8B needs 4.5 GB for its weights at 4-bit, 5.8 GB with an 8K context and 7.1 GB with 32K. At 8-bit, 9.1 GB.

What is the minimum VRAM for InternVL3 8B?

About 4.9 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 InternVL3 8B?

7.9B parameters: 14.8 GB of weights at 16-bit, 4.5 GB at 4-bit.

Can InternVL3 8B run on an RTX 4090?

Yes. At 8-bit it holds 32K of context and writes about 83 tokens a second.

More models