minimax M2

What is minimax M2?

minimax M2 is minimax's open-weights language model. Released in 2025-12, it ships in 1 size (230B) and is licensed minimax License (open-weight). The most popular use case is exceptional multilingual capabilities for code engineering.

VRAM and hardware

The 230B size fits in 96 GB of VRAM.

Size	Min VRAM (Q4_K_M)	Recommended VRAM	Tokens/sec on 3090
230B	130 GB	195 GB	~11 tok/s

How to run minimax M2 locally

Option 1: Ollama (simplest)

ollama pull minimax-m2
ollama run minimax-m2

Option 2: Mullama (production)

mullama pull minimax-m2
mullama run minimax-m2

Option 3: llama.cpp (CLI)

# Download a GGUF from Hugging Face (search "minimax-m2 gguf")
./llama-cli -m minimax-m2.Q4_K_M.gguf -p "Hello, AI!"

Option 4: Python with Mullama or llama-cpp-python

from mullama import Model, Context
model = Model.load("minimax-m2.Q4_K_M.gguf", n_gpu_layers=99)
ctx = Context(model, n_ctx=4096)
print(ctx.generate("Hello, AI!", 256))

What you can build with minimax M2

• multilingual
• code
• agentic
• chat

Hardware it fits on

• Apple Silicon (8GB): 230B in MLX
• Apple Silicon (16GB): 230B in MLX
• Apple Silicon (32GB): 230B in MLX or GGUF
• Apple Silicon (64GB+): 230B in MLX
• RTX 3090 (24GB): 230B at Q4_K_M
• RTX 4090 (24GB): same as 3090, ~30% faster
• RTX 5090 (32GB): 230B at Q8_0
• 2× RTX 6000 Ada (96GB total): 230B at Q4_K_M

Related models in the minimax family

• minimax M2.1 — 230B · December 2025 update to minimax M2
• minimax M2.5 — 230B · state-of-the-art real-world productivity and coding (Feb 2026)
• minimax M2.7 — 480B · minimax M2-series for coding, agentic workflows, professional productivity (Mar 2026)
• minimax M3 — 230B · latest minimax M-series (June 2026)

Other models you might consider

• Llama 3.1 (Meta) — 8B, general-purpose chat, agentic workflows, code
• Llama 3.2 (Meta) — 1B, low-end hardware, mobile, edge deployment
• Llama 3.2 Vision (Meta) — 11B, vision + chat on desktop GPUs
• Llama 3.3 (Meta) — 70B, best-in-class 70B for desktops
• Llama 4 (Meta) — Scout 17B (109B MoE), vision + long context (10M tokens)
• Qwen 2.5 (Alibaba) — 0.5B, strongest 7B-32B on consumer hardware