How to Run MiniMax M3.1 Flash Locally
Built with MiniMax M3.1-Flash. Converted from MiniMax's original weights on release day.
[!NOTE] Two kinds of files.
AD-*are Atomic Dynamic layouts: the routed experts carry the compression, attention, the shared expert and the embeddings stay high, and an importance matrix computed on M3.1 Flash's own chat format guides the rounding. The other files are plain llama.cpp quants without an importance matrix, published within hours of the weights.All of them load in stock llama.cpp and the apps built on it: standard
minimax-m3architecture, no runtime patch.
Pick a file
| File | Kind | Size on disk | Memory for the model | Fits |
|---|---|---|---|---|
AD-IQ2_XXS |
Atomic Dynamic | 112.2 GB | 104.5 GiB | 128 GB machines: a 128 GB Mac with the GPU limit raised (context up to about 32k), a DGX Spark, a PC with 128 GB of RAM and a GPU |
Q2_K |
plain | 155.9 GB | 145.2 GiB | 192 GB machines. Two bits without an importance matrix lose noticeably more; prefer AD-Q2_K at this size |
AD-Q2_K |
Atomic Dynamic | 161.6 GB | 150.5 GiB | 192 GB machines |
Q3_K_M |
plain | 203.5 GB | 189.5 GiB | 256 GB machines (a 256 GB Mac with the GPU limit raised) |
AD-IQ3_S |
Atomic Dynamic | 208.9 GB | 194.6 GiB | 256 GB machines |
AD-Q4_K |
Atomic Dynamic | 257.2 GB | 239.5 GiB | 384 GB machines and up, or several GPUs |
Q4_K_M |
plain | 258.0 GB | 240.3 GiB | 384 GB machines and up, or several GPUs |
Download one of them, for example:
hf download AtomicChat/MiniMax-M3.1-Flash-GGUF --include "Q3_K_M/*" --local-dir MiniMax-M3.1-Flash
MiniMax M3.1 Flash has 428B parameters, 23B of them active per token, so it needs a lot of memory but runs fast for its size. Add memory for the context: the attention cache takes about 0.14 MB per token, which is 1.1 GB at 8k context and 4.5 GB at 32k. Memory is in GiB (1 GiB = 1.074 GB), which is how GPUs and Macs report it; sizes on disk are in GB.
[!NOTE] On a Mac, macOS lets the GPU use only part of the unified memory by default. Raise the limit before loading, for example
sudo sysctl iogpu.wired_limit_mb=245000on a 256 GB Mac (resets on reboot).
Running it
llama-server -m MiniMax-M3.1-Flash/Q3_K_M/MiniMax-M3.1-Flash-Q3_K_M-00001-of-00005.gguf -ngl 99 -c 32768 -fa on --jinja
or straight from the hub: llama-server -hf AtomicChat/MiniMax-M3.1-Flash-GGUF:Q3_K_M -ngl 99 -c 32768 -fa on --jinja.
The chat template is inside every file; --jinja applies it. The files are split into parts of up to 45 GB;
point llama.cpp at the first part and it loads the rest.
- Sparse attention. M3.1 Flash uses MiniMax Sparse Attention. llama.cpp runs it only with flash attention
on (
-fa on) and without--kv-unified; otherwise it falls back to dense attention and warns about it. - Thinking. The model decides on its own when to think. Set the depth with
--chat-template-kwargs '{"reasoning_effort":"high"}'(low,medium,high,xhighormax). - Sampling. MiniMax recommends
--temp 1.0 --top-p 0.95. - Images.
mmproj-MiniMax-M3.1-Flash-F16.gguf(1.7 GB) is the vision encoder: add--mmproj mmproj-MiniMax-M3.1-Flash-F16.gguf. It is converted from the original weights; we have not tested images with it yet. - llama.cpp. MiniMax-M3 support landed in July 2026; these files were made and checked with
f5027fb21307(23 October 2026). - Checked: every file answered a chat turn through the built-in template before upload. Quality numbers against BF16 are being measured and will be added here.
How these files were made
MiniMax ships M3.1 Flash with its dense weights in MXFP8 and its routed experts in NVFP4 (trained for it).
llama.cpp's converter does not read this combination yet, so we first unpacked the weights to BF16: the MXFP8
values are exact in BF16, and the NVFP4 values carry a per-tensor scale that rounds at about 0.2% per weight.
From that BF16 copy, upstream convert_hf_to_gguf.py built the GGUF and llama-quantize built the files above.
One difference from the original: M3.1 Flash uses sparse attention in all 60 blocks, while llama.cpp's MiniMax-M3 code runs the first three blocks with dense attention, so their sparse-attention indexer weights are not in these files. While the context fits the blocks the indexer would pick (16 blocks of 128 tokens plus the local block, about 2,200 tokens), the result is the same; with a longer context those three blocks look at all of it instead of their usual selection.
The AD-* files come from the same BF16 GGUF. Their importance matrix comes from llama-imatrix over about 2.4M tokens
(every other piece of a 4.75M-token build of agentic, code, reasoning, 30-language and long-document text
rendered through M3.1 Flash's own chat template), computed on a Q8_0 copy of the model because the BF16 does
not fit our GPU machine. Every one of the 128 experts in every block received tokens. In each file the
routed-expert blocks with the highest activation energy (the last six) get one step more precision and
the coldest early blocks one step less.
The DSpark draft model that MiniMax ships for speculative decoding is not included.
License
These files are a derivative of MiniMax M3.1 Flash and come under the MiniMax Community License. It allows non-commercial use freely. Commercial use requires showing "Built with MiniMax M3.1-Flash" and a one-time notice to api@minimax.io, and written authorization from MiniMax above 20 million US dollars of yearly revenue; see the license for the full terms and the prohibited uses.
Quantized by AtomicChat.