Overview of Chinese AI companies and what you can do with it

Hey, Joey here.

There’s a weird shift happening in the AI world right now…

While everyone’s busy waiting on GPT‑5 or whatever OpenAI’s cooking next, China’s quietly dropping open-source models that are just as powerful, and in some cases, better.

I went deep down the rabbit hole this week to figure out what’s legit, what’s useful, and what you can actually do with these models.

But that’s not all:

📌 Resource: New AI architecture achieves 100x faster reasoning

📌 Video: I built a full funnel using only AI tools in 30 minutes (and yes, it actually converts)

📌 Deep Dive: Chinese open-source AI models are here—and they’re not playing around.

Let’s get into it.

WEEKLY PICKS

🗞️ Quick Reads:

  • New AI architecture delivers 100x faster reasoning (Article)

  • China latest AI cheaper than DeepSeek (Article)

👀 Video: I Build A Funnel Using Only AI in 30min (YouTube)

DEEP DIVE
Chinese AI Companies: what’s the fuss about

So you know how we always talk about ChatGPT, Claude, Grok, Gemini, yadda yadda?

Yeah, we’ve been sleeping on the non-US ones.

Because while the U.S. was flexing their AI models on stage... China was quietly cooking.

And now they’re coming out of the kitchen with heat.

Like actually scary-good models. Open-source. Run-them-locally-if-you-want type stuff.

And they’re almost FREE? Or at least way cheaper.

So I spent a few days deep in the rabbit hole, testing what the Chinese AI scene’s been building.

Let me show you what I found 👇

#1 GLM 4.5 – by Z AI

What it is:

An open-source AI chatbot, just like ChatGPT. You can use it in a browser, or download and run it locally. It’s designed for general conversation, coding help, reasoning, and even plugin-style tools.

How it compares:

This thing is fast and smart. It holds up well against GPT‑4. And since it’s open-source, you can tweak it to fit your use case or build your own assistant with it.

What it can do:

General chat, coding, tool use, research, agent-style workflows. Great for devs who want control.

U.S. equivalent:

Think GPT‑4o or Claude, but cheaper.

#2 Qwen 3 – by Alibaba Cloud

What it is:

A powerful language model (LLM) built for multiple use cases. You can run it as an AI chatbot, integrate it into apps, or use it behind the scenes for coding, summarizing, or even image-based tasks.

How it compares:

Qwen 3 is a legit challenger to Claude and Gemini, especially strong on coding and multilingual prompts. It also supports multimodal inputs like images and has text-to-speech support.

What it can do:

Conversations, document reading, coding, summarizing, multimodal analysis. Handles both Chinese and English effortlessly.

U.S. equivalent:

Gemini Pro or GPT‑4o—but easier to customize and self-host.

#3 Kimi K2 – by Moonshot AI

What it is:

A high-end AI chatbot app, similar to ChatGPT. It’s designed for long-form conversation, deep reasoning, complex questions, and creative tasks.

How it compares:

Kimi K2 absolutely crushes in areas like logic, code, and planning. It ranks at the top of many public benchmarks and even beat Claude and Gemini on several reasoning tasks.

What it can do:

Coding help, brainstorming, writing, answering multi-step questions. It’s like talking to an AI that can actually keep up with layered ideas.

U.S. equivalent:

Claude Opus or Grok, but faster and less filtered.

#4 DeepSeek – by DeepSeek AI

What it is:

A general-purpose AI model, open-source and easy to run. Built as a chatbot or backend assistant for anything from content to customer support.

How it compares:

It was trained on a shoestring budget, but still performs close to GPT‑4. Its efficiency is insane. It’s already being used in lightweight AI agents and apps.

What it can do:

Daily chat tasks, long-form writing, summarization, reasoning, automation triggers. Works great in DIY AI tools.

U.S. equivalent:

A more nimble GPT‑4 or early GPT‑5. Solid results with low compute.

#5 MiniMax M1 – by MiniMax AI

What it is:

An LLM built specifically for ultra-long context. It’s used as a chatbot or backend tool that can handle huge documents, think PDFs, research papers, transcripts and not forget halfway through.

How it compares:

This model is ideal for memory-intensive tasks. It’s one of the few open models that can consistently handle 1M tokens without breaking.

What it can do:

Summarize large files, reason across long chains of info, answer questions based on entire books. Also useful for memory-based agents.

U.S. equivalent:

GPT‑4 with long-context support (but no throttling or API costs).

#6 Wan 2.1 / 2.2 – by Alibaba

What it is:

An open-source text-to-video model. You type a prompt—like “robot doing tai chi at sunset”—and it generates a full HD video from it. It also supports image-to-video.

How it compares:

It’s fast. It’s surprisingly good. And it rivals tools like Runway Gen‑3 or even Meta’s Sora, but it’s open and accessible.

What it can do:

Generate videos for ads, content, storytelling, or creative projects. Handles bilingual prompts (Chinese + English) and renders in seconds.

U.S. equivalent:

Runway ML or OpenAI’s Sora—but with no waiting list or closed beta.

Quick Summary

Model

What it is

Best U.S. comparison

GLM 4.5

Open chatbot you can customize

GPT-4o / Claude

Qwen 3

Multimodal AI assistant

Gemini Pro + GPT-4

Kimi K2

Long-form, deep-thought bot

Claude Opus / Grok

DeepSeek

Budget-friendly chatbot

GPT-3.5 with sharp edges

MiniMax

AI for long docs + memory

GPT-4 with 1M token input

Wan 2.1

Text-to-video generator

Runway ML / Meta Sora

These models aren’t just cheap. They’re GOOD. Some of them feel better than what we’re paying for.

And most are open-source. You can download them. Tweak them. Build with them. No paywall. No tokens.

If you're an indie builder, agency, or just a nerd like me, you should be paying attention.

THAT‘S A WRAP

Before you go: Here’s how I can help

1) Sponsor Us — Reach 250,000+ AI enthusiasts, developers, and entrepreneurs monthly. Let’s collaborate →

2) The AI Content Machine Challenge — Join our 28-Day Generative AI Mastery Course. Master ChatGPT, Midjourney, and more with 5-minute daily challenges. Start creating 10x faster →

See you next week,

— Joey Mazars, Online Education & AI Expert 🥐

PS: Forward this to a friend who’s curious about AI. They’ll thank you (and so will I).

What'd you think of today's email?

Login or Subscribe to participate in polls.

Reply

or to participate.