MiniMax M2.7 — MiniMax 的本地 AI 模型
作者： Jakub Rusinowski · 最后更新：
MiniMax's March 2026 agentic MoE: 230B total parameters with 10B active per token across 256 experts — a 4.3% activation rate — over 62 layers, using multi-head causal self-attention with RoPE. Weights are published openly on Hugging Face and GitHub, so this is self-hostable rather than API-only, though the scale still implies a multi-GPU box. NVIDIA has upstreamed fused QK RMS Norm and FP8 MoE kernels from TensorRT-LLM for it. ⚠️ Released under a NON-COMMERCIAL license — commercial deployment requires a separate agreement with MiniMax.
变体
MiniMax M2.7 最小的变体在 Q4_K_M 下约需 140 GB 显存——量化权重加框架开销，不含 KV 缓存。
| 模型 | Q4 下显存 | 显存 | 上下文 | 运行 |
|---|---|---|---|---|
| MiniMax M2.7 230B-A10B → 230B (10B active) | ~139.7 GB | Unverified — sources disagree (192K–200K) | ollama run minimax-m2-7 |
显存为 Q4_K_M 下的量化权重加开销，与 GPU 与显存检测器使用同一引擎计算。
如何在本地运行 MiniMax M2.7
安装 Ollama，然后拉取标签。
ollama run minimax-m2-7在上方选择一个尺寸，查看它自己的显存、速度估算和安装命令。
许可证
Research / non-commercial only — this licence does NOT permit shipping a commercial product.
适用于: MiniMax M2.7 230B-A10B推荐 GPU
目录中能在本地运行 MiniMax M2.7（至少 140 GB 显存）的最便宜 GPU 是 Apple M3 Ultra (512 GB).
我的 GPU 能运行 MiniMax M2.7 吗？
- MiniMax M2.7 在 AMD Ryzen AI Max+ 395 上
- MiniMax M2.7 在 Apple M1 Max 上
- MiniMax M2.7 在 Apple M1 Ultra 上
- MiniMax M2.7 在 Apple M2 Max 上
- MiniMax M2.7 在 Apple M2 Ultra 上
- MiniMax M2.7 在 Apple M3 Max 上
- MiniMax M2.7 在 Apple M4 Max 上
- MiniMax M2.7 在 Apple M5 Max 上
- MiniMax M2.7 在 Apple M5 Pro 上
- MiniMax M2.7 在 NVIDIA A100 80GB (PCIe) 上
- MiniMax M2.7 在 NVIDIA DGX Spark 上
- MiniMax M2.7 在 NVIDIA H100 80GB (PCIe) 上
- MiniMax M2.7 在 NVIDIA RTX PRO 6000 Blackwell 上
MiniMax M2.7 — 常见问题
How much VRAM does MiniMax M2.7 need?
MiniMax M2.7 needs about 140 GB VRAM at Q4_K_M quantization for its smallest variant. Variants: MiniMax M2.7 230B-A10B (140 GB, Q4_K_M). On Apple Silicon, unified memory counts toward this requirement.
Can I run MiniMax M2.7 on an RTX 4090 (24 GB)?
MiniMax M2.7's smallest variant needs about 140 GB, which exceeds a single RTX 4090 (24 GB). Use multiple GPUs, a higher-VRAM card, or Apple Silicon with large unified memory.
What quantization should I use for MiniMax M2.7?
Q4_K_M is the best balance of quality and VRAM for MiniMax M2.7 in most cases. Choose Q8_0 for near-lossless quality if you have spare VRAM, or smaller quants (Q3/Q2) only when memory is tight.
How do I run MiniMax M2.7 with Ollama?
MiniMax M2.7 has no local Ollama tag — the published tag is cloud-hosted, so running it sends your prompts to a hosted GPU rather than your own machine.