Llama 4.5 — Meta 的本地 AI 模型

作者： Jakub Rusinowski · 最后更新：

Llama 4.5, reported as a June 2026 mid-cycle refresh of Llama 4 Scout and Maverick. UNVERIFIED: secondary sources disagree on whether this release happened, and no primary source has been read. Meta's confirmed 2026 open release is Muse Glimmer 30B (10 August, Apache-2.0).

Hugging Face Meta AI

变体

Llama 4.5 最小的变体在 Q4_K_M 下约需 67 GB 显存——量化权重加框架开销，不含 KV 缓存。

模型显存
Llama 4.5 Scout →
109B (17B active)
~66.6 GB

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 Llama 4.5

安装 Ollama，然后拉取标签。

ollama run llama-4-5

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

Llama Community允许商业使用

Weights are downloadable and commercial use is permitted, subject to the licence’s acceptable-use terms.

适用于: Llama 4.5 Scout

推荐 GPU

目录中能在本地运行 Llama 4.5（至少 67 GB 显存）的最便宜 GPU 是 NVIDIA DGX Spark (128 GB).

联盟营销声明: 本页部分链接为联盟推广链接——如果你通过它们购买，LLM Configurator 可能会获得佣金，而你无需支付任何额外费用。作为亚马逊联盟成员（Amazon Associate），LLM Configurator 会从符合条件的购买中获得收益。
NVIDIA DGX Spark (128GB)
128 GB VRAM · 150 W board power
2026年价格波动较大——请以当前商品页价格为准。

我的 GPU 能运行 Llama 4.5 吗？

Llama 4.5 — 常见问题

How much VRAM does Llama 4.5 need?

Llama 4.5 needs about 67 GB VRAM at Q4_K_M quantization for its smallest variant. Variants: Llama 4.5 Scout (67 GB, Q4_K_M). On Apple Silicon, unified memory counts toward this requirement.

Can I run Llama 4.5 on an RTX 4090 (24 GB)?

Llama 4.5's smallest variant needs about 67 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 Llama 4.5?

Q4_K_M is the best balance of quality and VRAM for Llama 4.5 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 Llama 4.5 with Ollama?

Llama 4.5 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.