Instructions to use Youssofal/MiniMax-M2.7-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Youssofal/MiniMax-M2.7-GGUF with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="Youssofal/MiniMax-M2.7-GGUF",
	filename="MiniMax-M2.7-IQ4_XS/MiniMax-M2.7-IQ4_XS-00001-of-00003.gguf",
)

llm.create_chat_completion(
	messages = [
		{
			"role": "user",
			"content": "What is the capital of France?"
		}
	]
)

Notebooks
Google Colab
Kaggle
Local Apps

llama.cpp

How to use Youssofal/MiniMax-M2.7-GGUF with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf Youssofal/MiniMax-M2.7-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf Youssofal/MiniMax-M2.7-GGUF:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf Youssofal/MiniMax-M2.7-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf Youssofal/MiniMax-M2.7-GGUF:Q4_K_M

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf Youssofal/MiniMax-M2.7-GGUF:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf Youssofal/MiniMax-M2.7-GGUF:Q4_K_M

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf Youssofal/MiniMax-M2.7-GGUF:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf Youssofal/MiniMax-M2.7-GGUF:Q4_K_M

Use Docker

docker model run hf.co/Youssofal/MiniMax-M2.7-GGUF:Q4_K_M

LM Studio
Jan

vLLM

How to use Youssofal/MiniMax-M2.7-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Youssofal/MiniMax-M2.7-GGUF"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Youssofal/MiniMax-M2.7-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Youssofal/MiniMax-M2.7-GGUF:Q4_K_M

Ollama
How to use Youssofal/MiniMax-M2.7-GGUF with Ollama:
```
ollama run hf.co/Youssofal/MiniMax-M2.7-GGUF:Q4_K_M
```

Unsloth Studio

How to use Youssofal/MiniMax-M2.7-GGUF with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for Youssofal/MiniMax-M2.7-GGUF to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for Youssofal/MiniMax-M2.7-GGUF to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for Youssofal/MiniMax-M2.7-GGUF to start chatting

How to use Youssofal/MiniMax-M2.7-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama-server -hf Youssofal/MiniMax-M2.7-GGUF:Q4_K_M

Configure the model in Pi

# Install Pi:
npm install -g @mariozechner/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "Youssofal/MiniMax-M2.7-GGUF:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use Youssofal/MiniMax-M2.7-GGUF with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama-server -hf Youssofal/MiniMax-M2.7-GGUF:Q4_K_M

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default Youssofal/MiniMax-M2.7-GGUF:Q4_K_M

Run Hermes

hermes

Docker Model Runner
How to use Youssofal/MiniMax-M2.7-GGUF with Docker Model Runner:
```
docker model run hf.co/Youssofal/MiniMax-M2.7-GGUF:Q4_K_M
```

Lemonade

How to use Youssofal/MiniMax-M2.7-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull Youssofal/MiniMax-M2.7-GGUF:Q4_K_M

Run and chat with the model

lemonade run user.MiniMax-M2.7-GGUF-Q4_K_M

List all available models

lemonade list

MiniMax-M2.7-GGUF

imatrix-calibrated GGUF quantizations of MiniMaxAI/MiniMax-M2.7 — the first model in the MiniMax lineup to participate in its own training evolution.

M2.7 matches or exceeds GPT-5.3-Codex on SWE-Pro (56.22%), rivals Opus 4.6 on VIBE-Pro (55.6%), and outperforms Claude Sonnet 4.6 on agentic benchmarks — with only 10B active parameters per token out of 229B total.

Why these quants

imatrix calibrated — all sub-Q8 quants use importance matrix for optimal quality retention. No blind quantization.
Q8_0 is effectively lossless — source weights are FP8 (float8_e4m3fn). Q8_0 preserves full precision.
XL variants — embed and output weights kept at Q8_0 for meaningfully better coherence on MoE routing. Available for Q4_K_M, Q5_K_M, Q3_K_M.
Correct license metadata — NON-COMMERCIAL license properly set in GGUF headers via --override-kv. Other repos incorrectly list modified-MIT.
Validated — every quant smoke-tested for coherent output, reasoning accuracy, and code generation before upload.

License

NON-COMMERCIAL USE ONLY. Commercial use requires prior written authorization from MiniMax. See LICENSE.

Quick Start

# Recommended: Q4_K_M for most users
hf download Youssofal/MiniMax-M2.7-GGUF --include "MiniMax-M2.7-Q4_K_M/*" --local-dir ./

# Run with llama-server
llama-server -m MiniMax-M2.7-Q4_K_M.gguf \
  -ngl 999 -c 32768 --jinja --reasoning-format auto -fa \
  --temp 1.0 --top-p 0.95 --top-k 40

Which file should I use?

Use Case	Recommended	Size	Notes
Maximum quality, unlimited RAM	Q8_0	~243 GB	Effectively lossless (FP8 source)
High quality, recommended	Q6_K	~188 GB	Near-perfect
Best quality/size balance	Q4_K_M	~138 GB	Default for most users
Better coherence at Q4	Q4_K_M-XL	~140 GB	Q8_0 embed/output weights
Low RAM / Apple Silicon 128GB	IQ4_XS	~122 GB	Best quality under 128 GB
Minimum viable	Q3_K_M	~109 GB	Usable but noticeable quality loss
Absolute minimum	Q2_K	~83 GB	Surprisingly usable

Sizing rule: Pick a quant 1-2 GB smaller than your total VRAM (GPU) or VRAM+RAM combined (Apple Silicon). A great deep-dive with charts is provided by Artefact2.

All Available Quantizations

Filename	Quant	Size	Split	Description
MiniMax-M2.7-Q8_0	Q8_0	~243 GB	Yes	Effectively lossless — source is FP8
MiniMax-M2.7-Q6_K	Q6_K	~188 GB	Yes	Very high quality, recommended
MiniMax-M2.7-Q5_K_M	Q5_K_M	~162 GB	Yes	High quality
MiniMax-M2.7-Q5_K_S	Q5_K_S	~157 GB	Yes	High quality
MiniMax-M2.7-Q4_K_M-XL	Q4_K_M	~140 GB	Yes	Q8_0 embed/output, better coherence
MiniMax-M2.7-Q4_K_M	Q4_K_M	~138 GB	Yes	Good quality, recommended default
MiniMax-M2.7-Q4_K_S	Q4_K_S	~130 GB	Yes	Slightly lower quality
MiniMax-M2.7-IQ4_XS	IQ4_XS	~122 GB	Yes	Best quality under 128 GB
MiniMax-M2.7-Q3_K_M-XL	Q3_K_M	~110 GB	Yes	Q8_0 embed/output, better than standard Q3
MiniMax-M2.7-Q3_K_L	Q3_K_L	~118 GB	Yes	Lower quality, usable
MiniMax-M2.7-Q3_K_M	Q3_K_M	~109 GB	Yes	Low quality
MiniMax-M2.7-Q3_K_S	Q3_K_S	~99 GB	Yes	Not recommended
MiniMax-M2.7-Q2_K	Q2_K	~83 GB	Yes	Very low quality, surprisingly usable

Download

pip install -U "huggingface_hub[cli]"

# Most quants are split (>50 GB) -- download the folder
hf download Youssofal/MiniMax-M2.7-GGUF --include "MiniMax-M2.7-Q4_K_M/*" --local-dir ./

# Smaller quants (single file)
hf download Youssofal/MiniMax-M2.7-GGUF --include "MiniMax-M2.7-Q2_K*" --local-dir ./

Running

# llama-server (recommended)
llama-server -m MiniMax-M2.7-Q4_K_M.gguf \
  -ngl 999 -c 32768 --jinja --reasoning-format auto -fa \
  --temp 1.0 --top-p 0.95 --top-k 40

# llama-cli (conversation mode)
llama-cli -m MiniMax-M2.7-Q4_K_M.gguf \
  -ngl 999 --jinja --reasoning-format auto -cnv

Sampling parameters (official MiniMax defaults): temperature 1.0, top_p 0.95, top_k 40.

Apple Silicon / MoE Notes

Despite only 10B active parameters per token, all 229B weights must reside in memory. On Apple Silicon with unified memory, use --cpu-moe for improved expert dispatch.

Q4_K_M (~138 GB) fits on 192 GB Mac
IQ4_XS (~122 GB) fits on 128 GB with swap pressure
Q3_K_M (~109 GB) runs on 128 GB Mac

Architecture

Spec	Value
Parameters	229B total, 10B active (4.3% activation)
Architecture	Sparse MoE, `minimax_m2`
Experts	256 local, 8 per token (top-k routing)
Layers	62
Attention	48 heads, 8 KV heads, hybrid Lightning + softmax
Context	200K (rope_theta 5,000,000)
Source precision	FP8 (float8_e4m3fn)
Thinking	`<think>...</think>` interleaved reasoning

XL Variants

XL variants (Q4_K_M-XL, Q5_K_M-XL, Q3_K_M-XL) keep embedding and output weight tensors at Q8_0 instead of the default lower precision. This preserves vocabulary quality at minimal size cost and meaningfully improves coherence -- particularly important for MoE models where expert routing precision matters.

Chat Template

Two templates are provided:

chat_template.official.jinja -- exact copy from MiniMax, includes <think> in generation prompt
chat_template.llama-server-patched.jinja -- removes <think> from generation prompt to prevent duplication in some llama-server versions

Compatibility

Engine	Status
llama.cpp (latest main)	Tested
LM Studio	Check for minimax_m2 support
Ollama	Requires recent version with minimax_m2
KoboldCpp	Untested

Quantization Details

Quantized with llama.cpp at commit ff5ef82
imatrix generated from bartowski calibration data
License metadata corrected via --override-kv (upstream model card incorrectly propagates modified-MIT)
All quants validated: coherent English output, correct reasoning, valid code generation

Credits

Original model by MiniMax -- MiniMaxAI/MiniMax-M2.7
imatrix calibration data by bartowski
Quantized by Youssofal

Downloads last month: 1,464

GGUF

Model size

229B params

Architecture

minimax-m2

Hardware compatibility

2-bit

4-bit

8-bit

Model tree for Youssofal/MiniMax-M2.7-GGUF

Base model

MiniMaxAI/MiniMax-M2.7

Quantized

(107)

this model