Skip to content

MiMo V2.5 VRAM requirements

How much VRAM MiMo V2.5 needs
4-bit weights: 175 GBWith 8K context: 177 GBWith 32K context: 177 GB
Parameters
311B
Active per token
17.4B
Longest context
1M
Released
Apr 2026
License
MIT
Strongest at
Not rated

How much VRAM MiMo V2.5 needs

Memory MiMo V2.5 needs at each precision and context length
PrecisionWeightsWith 8KWith 32KWith 128K
As released (FP8)293 GB294 GB295 GB297 GB
6-bit237 GB238 GB239 GB241 GB
5-bit206 GB207 GB207 GB210 GB
4-bit175 GB177 GB177 GB179 GB
3-bit141 GB143 GB143 GB145 GB
2-bit109 GB110 GB110 GB112 GB
1.6-bit59.7 GB60.8 GB61.3 GB63.4 GB

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

GPUs and Macs that run MiMo V2.5

The best way MiMo V2.5 runs on common GPUs, Macs and AI boxes
MachineBest fitContextSpeedFirst replyReport
RTX 4060Too bigPDF: MiMo V2.5 on RTX 4060 report
RTX 5060Too bigPDF: MiMo V2.5 on RTX 5060 report
RTX 3070Too bigPDF: MiMo V2.5 on RTX 3070 report
RTX 3080 10 GBToo bigPDF: MiMo V2.5 on RTX 3080 10 GB report
RTX 3060 12 GBToo bigPDF: MiMo V2.5 on RTX 3060 12 GB report
Arc B580Too bigPDF: MiMo V2.5 on Arc B580 report
RTX 4070Too bigPDF: MiMo V2.5 on RTX 4070 report
RTX 4070 SuperToo bigPDF: MiMo V2.5 on RTX 4070 Super report
RTX 5070Too bigPDF: MiMo V2.5 on RTX 5070 report
RTX 4060 Ti 16 GBToo bigPDF: MiMo V2.5 on RTX 4060 Ti 16 GB report
RTX 5060 Ti 16 GBToo bigPDF: MiMo V2.5 on RTX 5060 Ti 16 GB report
Arc A770 16 GBToo bigPDF: MiMo V2.5 on Arc A770 16 GB report
Radeon RX 9070 XTToo bigPDF: MiMo V2.5 on Radeon RX 9070 XT report
RTX 4070 Ti SuperToo bigPDF: MiMo V2.5 on RTX 4070 Ti Super report
RTX 4080 SuperToo bigPDF: MiMo V2.5 on RTX 4080 Super report
RTX 5070 TiToo bigPDF: MiMo V2.5 on RTX 5070 Ti report
RTX 5080Too bigPDF: MiMo V2.5 on RTX 5080 report
Radeon RX 7900 XTXToo bigPDF: MiMo V2.5 on Radeon RX 7900 XTX report
RTX 3090Too bigPDF: MiMo V2.5 on RTX 3090 report
RTX 4090Too bigPDF: MiMo V2.5 on RTX 4090 report
Radeon AI PRO R9700Too bigPDF: MiMo V2.5 on Radeon AI PRO R9700 report
RTX 5090Too bigPDF: MiMo V2.5 on RTX 5090 report
2× RTX 3090Too bigPDF: MiMo V2.5 on 2× RTX 3090 report
2× RTX 4090Too bigPDF: MiMo V2.5 on 2× RTX 4090 report
RTX PRO 5000 Blackwell 48 GBToo bigPDF: MiMo V2.5 on RTX PRO 5000 Blackwell 48 GB report
RTX 6000 AdaToo bigPDF: MiMo V2.5 on RTX 6000 Ada report
L40SToo bigPDF: MiMo V2.5 on L40S report
2× RTX 5090Too bigPDF: MiMo V2.5 on 2× RTX 5090 report
4× RTX 3090Too bigPDF: MiMo V2.5 on 4× RTX 3090 report
RTX PRO 6000 BlackwellToo bigPDF: MiMo V2.5 on RTX PRO 6000 Blackwell report
Mac mini M4 16 GBToo bigPDF: MiMo V2.5 on Mac mini M4 16 GB report
Mac mini M4 24 GBToo bigPDF: MiMo V2.5 on Mac mini M4 24 GB report
Mac mini M4 32 GBToo bigPDF: MiMo V2.5 on Mac mini M4 32 GB report
Mac mini M4 Pro 24 GBToo bigPDF: MiMo V2.5 on Mac mini M4 Pro 24 GB report
Mac mini M4 Pro 48 GBToo bigPDF: MiMo V2.5 on Mac mini M4 Pro 48 GB report
Mac mini M4 Pro 64 GBToo bigPDF: MiMo V2.5 on Mac mini M4 Pro 64 GB report
M4 Max 128 GBToo bigPDF: MiMo V2.5 on M4 Max 128 GB report
M5 Max 128 GBToo bigPDF: MiMo V2.5 on M5 Max 128 GB report
M3 Ultra 96 GBToo bigPDF: MiMo V2.5 on M3 Ultra 96 GB report
M3 Ultra 256 GB4-bit256K38 tok/s10 sPDF: MiMo V2.5 on M3 Ultra 256 GB report
M3 Ultra 512 GBFP81M25 tok/s10 sPDF: MiMo V2.5 on M3 Ultra 512 GB report
M5 Ultra 512 GBFP81M34 tok/s3.3 sPDF: MiMo V2.5 on M5 Ultra 512 GB report
Ryzen AI Max+ 395 64 GBToo bigPDF: MiMo V2.5 on Ryzen AI Max+ 395 64 GB report
Ryzen AI Max+ 395 128 GBToo bigPDF: MiMo V2.5 on Ryzen AI Max+ 395 128 GB report
RTX Spark2-bit224K29 tok/s3.7 sPDF: MiMo V2.5 on RTX Spark report
DGX Spark2-bit224K27 tok/s3.7 sPDF: MiMo V2.5 on DGX Spark report
Dual DGX Spark5-bit960K15 tok/s2.6 sPDF: MiMo V2.5 on Dual DGX Spark report
DGX Station GB3006-bit52K174 tok/s0.3 sPDF: MiMo V2.5 on DGX Station GB300 report
2× RTX PRO 6000 Blackwell4-bit240K88 tok/s0.5 sPDF: MiMo V2.5 on 2× RTX PRO 6000 Blackwell report
4× RTX PRO 6000 BlackwellFP81M59 tok/s0.3 sPDF: MiMo V2.5 on 4× RTX PRO 6000 Blackwell report
8× RTX PRO 6000 BlackwellFP81M59 tok/s0.2 sPDF: MiMo V2.5 on 8× RTX PRO 6000 Blackwell report
A100 80 GB SXMToo bigPDF: MiMo V2.5 on A100 80 GB SXM report
H100 SXMToo bigPDF: MiMo V2.5 on H100 SXM report
4× H200 NVLFP81M123 tok/s0.3 sPDF: MiMo V2.5 on 4× H200 NVL report
8× H200 NVLFP81M123 tok/s0.1 sPDF: MiMo V2.5 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/MiMo-V2.5-GGUF
vLLM
vllm serve XiaomiMiMo/MiMo-V2.5

Questions

How much VRAM does MiMo V2.5 need?

MiMo V2.5 needs 175 GB for its weights at 4-bit, 177 GB with an 8K context and 177 GB with 32K.

What is the minimum VRAM for MiMo V2.5?

About 110 GB at 2-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 M3 Ultra 256 GB.

How big is MiMo V2.5?

311B parameters (17.4B active per token): 293 GB of weights as released (FP8), 175 GB at 4-bit.

Can MiMo V2.5 run on an RTX 4090?

Not on one card. The smallest machine here that runs it is the M3 Ultra 256 GB.

More models