ai-agent

kimi-k3-mlx

Name: kimi-k3-mlx
Author: PipeNetwork

MLX port of moonshotai/Kimi-K3 (2.78T multimodal MoE): streaming converter, REAP expert pruning, and per-language expert-overlap analysis

340 stars40 forksPythonUpdated 8/1/2026100% free · open source

What it does

The kimi-k3-mlx tool is a streaming converter, REAP expert pruning, and per-language expert-overlap analysis for the 2.78T multimodal MoE model, allowing for efficient processing and analysis of large datasets

When to use it

•When you need to process and analyze large multimodal datasets
•When you want to prune experts for more efficient model deployment
•When you need to perform per-language expert-overlap analysis

Ready-to-paste prompt

python -m kimi_k3_mlx.convert --input-path /path/to/input --output-path /path/to/output

Heads up: Make sure you have the correct version of Python and the required dependencies installed, as specified in the repository's README, to avoid compatibility issues

Saves to your device

Use with Claude

New

Skip the builder — one click puts this in Claude, Cursor, Antigravity and more.

✅ Light setup

Installs with a command or two; your AI agent can do it for you.

Try it instantly — no install

Claude Code

mkdir -p ~/.claude/skills/kimi-k3-mlx && curl -fsSL https://workflowstacks.com/api/skills/kimi-k3-mlx/claude-skill -o ~/.claude/skills/kimi-k3-mlx/SKILL.md

Open in another AI app

Opens the app with this repo with the prompt ready to go — no copy-paste needed.

Connect the whole catalog (MCP)

claude mcp add --transport http workflowstacks https://workflowstacks.com/api/mcp

Adds a WorkflowStacks connector to Claude Code: search and load any skill here by chatting.

How kimi-k3-mlx works

Codeflow

Free to inspect

Kimi-k3-mlx is a medium Python project (~11k lines across 45 code files, plus 10 test files). Setup is light: installs like a normal app. Reading the code is optional. Last commit 3 months ago, no license file, has a test suite.

Size

Medium codebase

~11k lines · 45 code files · ~2 h to skim

Setup

One-command install

Installs like a normal app. Reading the code is optional.

Runs on

Python

No API keys detected

Python 98%Shell 2%

Where to start reading

1
README.md
Start here — what it does and how to install it

What's in each folder

scripts/Helper scripts34 files

reference/Folder8 files

kimi_k3_vl/Folder5 files

tests/Tests — proof it works10 files

READMEHas testsThin docsNo CINo licenseUpdated 3 mo ago