Skip to content

EXAONE 4.0 1.2B VRAM requirements

How much VRAM EXAONE 4.0 1.2B needs
4-bit weights: 740 MBWith 8K context: 1.9 GBWith 32K context: 3.3 GB8-bit, 8K context: 2.4 GB
Parameters
1.3B
Longest context
64K
Released
Jul 2025
License
Exaone
Strongest at
Not rated

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

How much VRAM EXAONE 4.0 1.2B needs

Memory EXAONE 4.0 1.2B needs at each precision and context length
PrecisionWeightsWith 8KWith 32KWith 64K
16-bit2.4 GB3.6 GB5.0 GB6.8 GB
8-bit1.3 GB2.4 GB3.8 GB5.7 GB
6-bit1.0 GB2.2 GB3.6 GB5.4 GB
5-bit868 MB2.0 GB3.4 GB5.3 GB
4-bit740 MB1.9 GB3.3 GB5.2 GB
3-bit596 MB1.8 GB3.2 GB5.0 GB
2-bit458 MB1.6 GB3.0 GB4.9 GB
1.6-bit252 MB1.4 GB2.8 GB4.7 GB

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

GPUs and Macs that run EXAONE 4.0 1.2B

The best way EXAONE 4.0 1.2B runs on common GPUs, Macs and AI boxes
MachineBest fitContextSpeedFirst replyReport
RTX 40608-bit64K124 tok/s0.2 sPDF: EXAONE 4.0 1.2B on RTX 4060 report
RTX 50608-bit64K191 tok/s0.1 sPDF: EXAONE 4.0 1.2B on RTX 5060 report
RTX 30708-bit64K191 tok/s0.1 sPDF: EXAONE 4.0 1.2B on RTX 3070 report
RTX 3080 10 GB8-bit64K289 tok/s0.1 sPDF: EXAONE 4.0 1.2B on RTX 3080 10 GB report
RTX 3060 12 GB8-bit64K159 tok/s0.2 sPDF: EXAONE 4.0 1.2B on RTX 3060 12 GB report
Arc B5808-bit64K129 tok/s0.1 sPDF: EXAONE 4.0 1.2B on Arc B580 report
RTX 40708-bit64K210 tok/s0.1 sPDF: EXAONE 4.0 1.2B on RTX 4070 report
RTX 4070 Super8-bit64K210 tok/s0.1 sPDF: EXAONE 4.0 1.2B on RTX 4070 Super report
RTX 50708-bit64K264 tok/s0.1 sPDF: EXAONE 4.0 1.2B on RTX 5070 report
RTX 4060 Ti 16 GB8-bit64K131 tok/s0.1 sPDF: EXAONE 4.0 1.2B on RTX 4060 Ti 16 GB report
RTX 5060 Ti 16 GB8-bit64K191 tok/s0.1 sPDF: EXAONE 4.0 1.2B on RTX 5060 Ti 16 GB report
Arc A770 16 GB8-bit64K150 tok/s0.1 sPDF: EXAONE 4.0 1.2B on Arc A770 16 GB report
Radeon RX 9070 XT8-bit64K198 tok/s0.1 sPDF: EXAONE 4.0 1.2B on Radeon RX 9070 XT report
RTX 4070 Ti Super8-bit64K264 tok/s0.1 sPDF: EXAONE 4.0 1.2B on RTX 4070 Ti Super report
RTX 4080 Super8-bit64K283 tok/s0.1 sPDF: EXAONE 4.0 1.2B on RTX 4080 Super report
RTX 5070 Ti8-bit64K326 tok/s0.1 sPDF: EXAONE 4.0 1.2B on RTX 5070 Ti report
RTX 50808-bit64K342 tok/s0.1 sPDF: EXAONE 4.0 1.2B on RTX 5080 report
Radeon RX 7900 XTX8-bit64K259 tok/s0.1 sPDF: EXAONE 4.0 1.2B on Radeon RX 7900 XTX report
RTX 30908-bit64K336 tok/s0.1 sPDF: EXAONE 4.0 1.2B on RTX 3090 report
RTX 40908-bit64K354 tok/s0.1 sPDF: EXAONE 4.0 1.2B on RTX 4090 report
Radeon AI PRO R97008-bit64K198 tok/s0.1 sPDF: EXAONE 4.0 1.2B on Radeon AI PRO R9700 report
RTX 50908-bit64K504 tok/s0.1 sPDF: EXAONE 4.0 1.2B on RTX 5090 report
2× RTX 30908-bit64K336 tok/s0.1 sPDF: EXAONE 4.0 1.2B on 2× RTX 3090 report
2× RTX 40908-bit64K354 tok/s0.1 sPDF: EXAONE 4.0 1.2B on 2× RTX 4090 report
RTX PRO 5000 Blackwell 48 GB8-bit64K426 tok/s0.1 sPDF: EXAONE 4.0 1.2B on RTX PRO 5000 Blackwell 48 GB report
RTX 6000 Ada8-bit64K342 tok/s0.1 sPDF: EXAONE 4.0 1.2B on RTX 6000 Ada report
L40S8-bit64K318 tok/s0.1 sPDF: EXAONE 4.0 1.2B on L40S report
2× RTX 50908-bit64K504 tok/s0.1 sPDF: EXAONE 4.0 1.2B on 2× RTX 5090 report
4× RTX 30908-bit64K336 tok/s0.1 sPDF: EXAONE 4.0 1.2B on 4× RTX 3090 report
RTX PRO 6000 Blackwell8-bit64K504 tok/s0.1 sPDF: EXAONE 4.0 1.2B on RTX PRO 6000 Blackwell report
Mac mini M4 16 GB8-bit64K62 tok/s1.6 sPDF: EXAONE 4.0 1.2B on Mac mini M4 16 GB report
Mac mini M4 24 GB8-bit64K62 tok/s1.6 sPDF: EXAONE 4.0 1.2B on Mac mini M4 24 GB report
Mac mini M4 32 GB8-bit64K62 tok/s1.6 sPDF: EXAONE 4.0 1.2B on Mac mini M4 32 GB report
Mac mini M4 Pro 24 GB8-bit64K118 tok/s0.8 sPDF: EXAONE 4.0 1.2B on Mac mini M4 Pro 24 GB report
Mac mini M4 Pro 48 GB8-bit64K118 tok/s0.8 sPDF: EXAONE 4.0 1.2B on Mac mini M4 Pro 48 GB report
Mac mini M4 Pro 64 GB8-bit64K118 tok/s0.8 sPDF: EXAONE 4.0 1.2B on Mac mini M4 Pro 64 GB report
M4 Max 128 GB8-bit64K175 tok/s0.4 sPDF: EXAONE 4.0 1.2B on M4 Max 128 GB report
M5 Max 128 GB8-bit64K187 tok/s0.1 sPDF: EXAONE 4.0 1.2B on M5 Max 128 GB report
M3 Ultra 96 GB8-bit64K202 tok/s0.2 sPDF: EXAONE 4.0 1.2B on M3 Ultra 96 GB report
M3 Ultra 256 GB8-bit64K202 tok/s0.2 sPDF: EXAONE 4.0 1.2B on M3 Ultra 256 GB report
M3 Ultra 512 GB8-bit64K202 tok/s0.2 sPDF: EXAONE 4.0 1.2B on M3 Ultra 512 GB report
M5 Ultra 512 GB8-bit64K241 tok/s0.1 sPDF: EXAONE 4.0 1.2B on M5 Ultra 512 GB report
Ryzen AI Max+ 395 64 GB8-bit64K97 tok/s0.2 sPDF: EXAONE 4.0 1.2B on Ryzen AI Max+ 395 64 GB report
Ryzen AI Max+ 395 128 GB8-bit64K97 tok/s0.2 sPDF: EXAONE 4.0 1.2B on Ryzen AI Max+ 395 128 GB report
RTX Spark8-bit64K136 tok/s0.1 sPDF: EXAONE 4.0 1.2B on RTX Spark report
DGX Spark8-bit64K125 tok/s0.1 sPDF: EXAONE 4.0 1.2B on DGX Spark report
Dual DGX Spark8-bit64K125 tok/s0.1 sPDF: EXAONE 4.0 1.2B on Dual DGX Spark report
DGX Station GB3008-bit64K852 tok/s0.1 sPDF: EXAONE 4.0 1.2B on DGX Station GB300 report
2× RTX PRO 6000 Blackwell8-bit64K504 tok/s0.1 sPDF: EXAONE 4.0 1.2B on 2× RTX PRO 6000 Blackwell report
4× RTX PRO 6000 Blackwell8-bit64K504 tok/s0.1 sPDF: EXAONE 4.0 1.2B on 4× RTX PRO 6000 Blackwell report
8× RTX PRO 6000 Blackwell8-bit64K504 tok/s0.1 sPDF: EXAONE 4.0 1.2B on 8× RTX PRO 6000 Blackwell report
A100 80 GB SXM8-bit64K540 tok/s0.1 sPDF: EXAONE 4.0 1.2B on A100 80 GB SXM report
H100 SXM8-bit64K676 tok/s0.1 sPDF: EXAONE 4.0 1.2B on H100 SXM report
4× H200 NVL8-bit64K767 tok/s0.1 sPDF: EXAONE 4.0 1.2B on 4× H200 NVL report
8× H200 NVL8-bit64K767 tok/s0.1 sPDF: EXAONE 4.0 1.2B 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 LGAI-EXAONE/EXAONE-4.0-1.2B-GGUF
vLLM
vllm serve LGAI-EXAONE/EXAONE-4.0-1.2B
MLX (Mac)
mlx_lm.chat --model mlx-community/exaone-4.0-1.2b-4bit

Questions

How much VRAM does EXAONE 4.0 1.2B need?

EXAONE 4.0 1.2B needs 740 MB for its weights at 4-bit, 1.9 GB with an 8K context and 3.3 GB with 32K. At 8-bit, 2.4 GB.

What is the minimum VRAM for EXAONE 4.0 1.2B?

About 1.8 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 EXAONE 4.0 1.2B?

1.3B parameters: 2.4 GB of weights at 16-bit, 740 MB at 4-bit.

Can EXAONE 4.0 1.2B run on an RTX 4090?

Yes. At 8-bit it holds 64K of context and writes about 354 tokens a second.

More models