Skip to content

How much VRAM does MiniMax M2.7 need?

MiniMax M2.7 needs 129 GB for its weights at 4-bit, 132 GB with an 8K context and 138 GB with 32K.

Parameters
229B
Active per token
10B
Longest context
200K
Released
Apr 2026
License
MiniMax Non-Commercial
Strongest at
Coding, Agents

Memory at each precision

Memory MiniMax M2.7 needs at each precision and context length
PrecisionWeightsWith 8KWith 32KWith 128K
As released (FP8)214 GB217 GB223 GB246 GB
6-bit175 GB178 GB183 GB207 GB
5-bit151 GB154 GB160 GB183 GB
4-bit129 GB132 GB138 GB161 GB
3-bit104 GB107 GB113 GB136 GB
2-bit79.9 GB82.7 GB88.5 GB112 GB
1.6-bit43.9 GB46.8 GB52.6 GB75.9 GB

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

What runs it

The best way MiniMax M2.7 runs on common GPUs, Macs and AI boxes
MachineBest fitContextSpeedFirst reply
RTX 3060 12 GBToo big
RTX 4060 Ti 16 GBToo big
RTX 3090Too big
RTX 4090Too big
RTX 5090Too big
2× RTX 3090Too big
RTX PRO 6000 Blackwell2-bit40K149 tok/s0.5 s
M4 Max 128 GB2-bit40K52 tok/s10 s
M5 Max 128 GB2-bit40K55 tok/s3.0 s
Ryzen AI Max+ 395 128 GB2-bit40K33 tok/s6.0 s
DGX Spark 128 GB3-bit36K34 tok/s2.3 s
M3 Ultra 256 GB6-bit28K42 tok/s5.9 s
M5 Ultra 512 GBFP8200K46 tok/s1.8 s
H100 SXMToo big
DGX Station GB300FP896K174 tok/s0.2 s
8× H200 NVLFP8200K149 tok/s0.1 s

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/MiniMax-M2.7-GGUF
vLLM
vllm serve MiniMaxAI/MiniMax-M2.7
MLX (Mac)
mlx_lm.chat --model mlx-community/MiniMax-M2.7-4bit

Questions

How much VRAM does MiniMax M2.7 need?

MiniMax M2.7 needs 129 GB for its weights at 4-bit, 132 GB with an 8K context and 138 GB with 32K.

Can MiniMax M2.7 run on an RTX 4090?

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

More models