Skip to content

Qwen3.8 2.4T A95B VRAM requirements

How much VRAM Qwen3.8 2.4T A95B needs
4-bit weights: 1,381 GBWith 8K context: 1,384 GBWith 32K context: 1,386 GB8-bit, 8K context: 2,423 GB
Parameters
2.4T
Active per token
95B
Longest context
256K
Released
Aug 2026
License
Qwen3.8-Max License
Strongest at
Chat, Coding, Agents

How much VRAM Qwen3.8 2.4T A95B needs

Memory Qwen3.8 2.4T A95B needs at each precision and context length
PrecisionWeightsWith 8KWith 32KWith 128K
16-bit4,556 GB4,559 GB4,561 GB4,570 GB
8-bit2,421 GB2,423 GB2,425 GB2,434 GB
6-bit1,868 GB1,871 GB1,873 GB1,881 GB
5-bit1,620 GB1,623 GB1,625 GB1,634 GB
4-bit1,381 GB1,384 GB1,386 GB1,394 GB
3-bit1,113 GB1,116 GB1,118 GB1,127 GB
2-bit854 GB857 GB859 GB868 GB
1.6-bit470 GB472 GB474 GB483 GB

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

GPUs and Macs that run Qwen3.8 2.4T A95B

The best way Qwen3.8 2.4T A95B runs on common GPUs, Macs and AI boxes
MachineBest fitContextSpeedFirst replyReport
RTX 4060Too bigPDF: Qwen3.8 2.4T A95B on RTX 4060 report
RTX 5060Too bigPDF: Qwen3.8 2.4T A95B on RTX 5060 report
RTX 3070Too bigPDF: Qwen3.8 2.4T A95B on RTX 3070 report
RTX 3080 10 GBToo bigPDF: Qwen3.8 2.4T A95B on RTX 3080 10 GB report
RTX 3060 12 GBToo bigPDF: Qwen3.8 2.4T A95B on RTX 3060 12 GB report
Arc B580Too bigPDF: Qwen3.8 2.4T A95B on Arc B580 report
RTX 4070Too bigPDF: Qwen3.8 2.4T A95B on RTX 4070 report
RTX 4070 SuperToo bigPDF: Qwen3.8 2.4T A95B on RTX 4070 Super report
RTX 5070Too bigPDF: Qwen3.8 2.4T A95B on RTX 5070 report
RTX 4060 Ti 16 GBToo bigPDF: Qwen3.8 2.4T A95B on RTX 4060 Ti 16 GB report
RTX 5060 Ti 16 GBToo bigPDF: Qwen3.8 2.4T A95B on RTX 5060 Ti 16 GB report
Arc A770 16 GBToo bigPDF: Qwen3.8 2.4T A95B on Arc A770 16 GB report
Radeon RX 9070 XTToo bigPDF: Qwen3.8 2.4T A95B on Radeon RX 9070 XT report
RTX 4070 Ti SuperToo bigPDF: Qwen3.8 2.4T A95B on RTX 4070 Ti Super report
RTX 4080 SuperToo bigPDF: Qwen3.8 2.4T A95B on RTX 4080 Super report
RTX 5070 TiToo bigPDF: Qwen3.8 2.4T A95B on RTX 5070 Ti report
RTX 5080Too bigPDF: Qwen3.8 2.4T A95B on RTX 5080 report
Radeon RX 7900 XTXToo bigPDF: Qwen3.8 2.4T A95B on Radeon RX 7900 XTX report
RTX 3090Too bigPDF: Qwen3.8 2.4T A95B on RTX 3090 report
RTX 4090Too bigPDF: Qwen3.8 2.4T A95B on RTX 4090 report
Radeon AI PRO R9700Too bigPDF: Qwen3.8 2.4T A95B on Radeon AI PRO R9700 report
RTX 5090Too bigPDF: Qwen3.8 2.4T A95B on RTX 5090 report
2× RTX 3090Too bigPDF: Qwen3.8 2.4T A95B on 2× RTX 3090 report
2× RTX 4090Too bigPDF: Qwen3.8 2.4T A95B on 2× RTX 4090 report
RTX PRO 5000 Blackwell 48 GBToo bigPDF: Qwen3.8 2.4T A95B on RTX PRO 5000 Blackwell 48 GB report
RTX 6000 AdaToo bigPDF: Qwen3.8 2.4T A95B on RTX 6000 Ada report
L40SToo bigPDF: Qwen3.8 2.4T A95B on L40S report
2× RTX 5090Too bigPDF: Qwen3.8 2.4T A95B on 2× RTX 5090 report
4× RTX 3090Too bigPDF: Qwen3.8 2.4T A95B on 4× RTX 3090 report
RTX PRO 6000 BlackwellToo bigPDF: Qwen3.8 2.4T A95B on RTX PRO 6000 Blackwell report
Mac mini M4 16 GBToo bigPDF: Qwen3.8 2.4T A95B on Mac mini M4 16 GB report
Mac mini M4 24 GBToo bigPDF: Qwen3.8 2.4T A95B on Mac mini M4 24 GB report
Mac mini M4 32 GBToo bigPDF: Qwen3.8 2.4T A95B on Mac mini M4 32 GB report
Mac mini M4 Pro 24 GBToo bigPDF: Qwen3.8 2.4T A95B on Mac mini M4 Pro 24 GB report
Mac mini M4 Pro 48 GBToo bigPDF: Qwen3.8 2.4T A95B on Mac mini M4 Pro 48 GB report
Mac mini M4 Pro 64 GBToo bigPDF: Qwen3.8 2.4T A95B on Mac mini M4 Pro 64 GB report
M4 Max 128 GBToo bigPDF: Qwen3.8 2.4T A95B on M4 Max 128 GB report
M5 Max 128 GBToo bigPDF: Qwen3.8 2.4T A95B on M5 Max 128 GB report
M3 Ultra 96 GBToo bigPDF: Qwen3.8 2.4T A95B on M3 Ultra 96 GB report
M3 Ultra 256 GBToo bigPDF: Qwen3.8 2.4T A95B on M3 Ultra 256 GB report
M3 Ultra 512 GBToo bigPDF: Qwen3.8 2.4T A95B on M3 Ultra 512 GB report
M5 Ultra 512 GBToo bigPDF: Qwen3.8 2.4T A95B on M5 Ultra 512 GB report
Ryzen AI Max+ 395 64 GBToo bigPDF: Qwen3.8 2.4T A95B on Ryzen AI Max+ 395 64 GB report
Ryzen AI Max+ 395 128 GBToo bigPDF: Qwen3.8 2.4T A95B on Ryzen AI Max+ 395 128 GB report
RTX SparkToo bigPDF: Qwen3.8 2.4T A95B on RTX Spark report
DGX SparkToo bigPDF: Qwen3.8 2.4T A95B on DGX Spark report
Dual DGX SparkToo bigPDF: Qwen3.8 2.4T A95B on Dual DGX Spark report
DGX Station GB3001.6-bit, spills256K22 tok/s4.5 sPDF: Qwen3.8 2.4T A95B on DGX Station GB300 report
2× RTX PRO 6000 BlackwellToo bigPDF: Qwen3.8 2.4T A95B on 2× RTX PRO 6000 Blackwell report
4× RTX PRO 6000 BlackwellToo bigPDF: Qwen3.8 2.4T A95B on 4× RTX PRO 6000 Blackwell report
8× RTX PRO 6000 Blackwell1.6-bit256K46 tok/s0.8 sPDF: Qwen3.8 2.4T A95B on 8× RTX PRO 6000 Blackwell report
A100 80 GB SXMToo bigPDF: Qwen3.8 2.4T A95B on A100 80 GB SXM report
H100 SXMToo bigPDF: Qwen3.8 2.4T A95B on H100 SXM report
4× H200 NVL1.6-bit256K87 tok/s1.5 sPDF: Qwen3.8 2.4T A95B on 4× H200 NVL report
8× H200 NVL2-bit256K61 tok/s0.8 sPDF: Qwen3.8 2.4T A95B 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 unsloth/Qwen3.8-2.4T-A95B-GGUF
vLLM
vllm serve Qwen/Qwen3.8-2.4T-A95B

Questions

How much VRAM does Qwen3.8 2.4T A95B need?

Qwen3.8 2.4T A95B needs 1,381 GB for its weights at 4-bit, 1,384 GB with an 8K context and 1,386 GB with 32K. At 8-bit, 2,423 GB.

What is the minimum VRAM for Qwen3.8 2.4T A95B?

About 472 GB at 1.6-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 8× RTX PRO 6000 Blackwell.

How big is Qwen3.8 2.4T A95B?

2.4T parameters (95B active per token): 4,556 GB of weights at 16-bit, 1,381 GB at 4-bit.

Can Qwen3.8 2.4T A95B run on an RTX 4090?

Not on one card. The smallest machine here that runs it is the 8× RTX PRO 6000 Blackwell.

More models