Mistral Medium 3.5 — Mistral AI 的本地 AI 模型
作者： Jakub Rusinowski · 最后更新：
Mistral 首个合并式旗舰：一个 128B 稠密模型，把推理线（Magistral）和编程线（Devstral 2）合并进同一个检查点，支持图像输入和 256K 上下文。稠密意味着每个 token 都要读取全部参数，所以每 token 比同等规模的混合专家模型慢，但 Q4_K_M 约 78 GB，可放进 128 GB 的 Mac 或两张 48 GB 显卡。它的“Modified MIT”许可证对月收入超过 2000 万美元的公司不授予任何权利。
变体
Mistral Medium 3.5 最小的变体在 Q4_K_M 下约需 78 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 |
|---|---|---|---|---|
| Mistral Medium 3.5 128B → 128B | ~77.9 GB | 262,144 | ollama run mistral-medium-3.5:128b |
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 Mistral Medium 3.5
安装 Ollama，然后拉取标签。
ollama run mistral-medium-3.5:128b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
Weights are downloadable and commercial use is permitted, subject to the licence’s acceptable-use terms.
适用于: Mistral Medium 3.5 128B推荐 GPU
目录中能在本地运行 Mistral Medium 3.5（至少 78 GB 显存）的最便宜 GPU 是 NVIDIA DGX Spark (128 GB).
Mistral Medium 3.5 — 常见问题
How much VRAM does Mistral Medium 3.5 need?
Mistral Medium 3.5 needs about 78 GB VRAM at Q4_K_M quantization for its smallest variant. Variants: Mistral Medium 3.5 128B (78 GB, Q4_K_M). On Apple Silicon, unified memory counts toward this requirement.
Can I run Mistral Medium 3.5 on an RTX 4090 (24 GB)?
Mistral Medium 3.5's smallest variant needs about 78 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 Mistral Medium 3.5?
Q4_K_M is the best balance of quality and VRAM for Mistral Medium 3.5 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 Mistral Medium 3.5 with Ollama?
Install Ollama, then run: ollama run mistral-medium-3.5:128b. This downloads Mistral Medium 3.5 and starts a local, OpenAI-compatible endpoint — no internet connection is needed after the initial download.