MiniMax-H3 全模态模型登陆苹果M系列芯片，MLX开源移植可本地跑

PipeNetwork/minimax-h3-mlx

PipeNetwork/minimax-h3-mlx。MiniMax 两天前发布了 MiniMax-H3——他们将其描述为"一个通用、全模态生成系统"，实际上意味着它能接受文本、图片、音频和视频，并能用它们生成最长 15 秒、包含音频的视频片段。

这个 Python 包将其移植到了 MLX 上，以便在 Apple Silicon 上运行。

我在 M5 Max MacBook Pro 上跑通了它。克隆仓库后，我是这样运行模型的：

# First download the models
uvx --from huggingface_hub hf download MiniMaxAI/MiniMax-H3 \
  --include 'FL2VA/*' --exclude 'FL2VA/transformer/*'
uvx --from huggingface_hub hf download pipenetwork/MiniMax-H3-MLX-8bit

# Now run the prompt
uv run --with mlx-vlm \
  --with-requirements requirements.txt python scripts/generate.py \
  "a rainbow colored skunk leaps over a mossy log in a supermarket" \
  -o skunk.mp4 \
  -c ~/.cache/huggingface/hub/models--MiniMaxAI--MiniMax-H3/snapshots/fa9c8ab1eaa21c8ae25e7e40b83b2e6002f340af/FL2VA \
  -t ~/.cache/huggingface/hub/models--pipenetwork--MiniMax-H3-MLX-8bit/snapshots/3ac52081470b0488921c3ec3ba84a39097bf2361

以下是这条 prompt 生成出来的视频：

a rainbow colored skunk leaps over a mossy log in a supermarket

Your browser does not support HTML5 video.

它下载了约 115 GB 的模型文件，生成视频耗时不到 45 分钟。

视频效果很惊艳，但音频听起来像是怪异的语音垃圾——因为我没有在 prompt 里给出关于音频应该是什么样的引导。Prompt 指南（在这次实验前我没看过）里有大量信息告诉你如何正确使用音频功能。

Stateless MCP has recaptured my interest (and inspired mcp-explorer and datasette-mcp) - 31st July 2026

OpenAI's accidental cyberattack against Hugging Face is science fiction that happened - 22nd July 2026

A Fireside Chat with Cat and Thariq from the Claude Code team - 21st July 2026

This is a link post by Simon Willison, posted on 4th August 2026.

Sponsor me for $10/month and get a curated email digest of the month's most important LLM developments.

Pay me to send you less!