5 Chinese AI Startups to Watch Beyond DeepSeek in 2025

China’s AI industry is growing quickly, with many new startups making progress beyond big names such as DeepSeek. While larger companies get most of the attention, smaller Chinese AI startups also create smart AI tools and advanced systems for businesses. With strong support and investment, they can compete worldwide.

Ben James, founder of 404, Bittensor Subnet 17, told Techopedia:

“DeepSeek has definitely drawn attention to China’s impressive AI ecosystem. While the work that DeepSeek is doing is impressive, there are more startups that are making waves while solving real-world problems.”

So, who are they? Techopedia looks at five of the most promising Chinese AI startups in 2025.

Table of Contents Table of Contents

1. Key Takeaways
2. Top 5 Chinese AI Companies to Watch in 2025
3. The Bottom Line
4. FAQs
5. References

Show Full Guide

Table of Contents

1. Key Takeaways
2. Top 5 Chinese AI Companies to Watch in 2025
3. The Bottom Line
Show Full Guide
4. FAQs
5. References

Top 5 Chinese AI Companies to Watch in 2025

1. Stepfun

Stepfun, an AI startup based in Shanghai, was founded in April 2023 by Jiang Daxin, the former chief scientist of the Microsoft Asia Research Institute. Stepfun is becoming a significant player in the AI market by building advanced Chinese AI models with trillion parameters and a mixture of experts (MoE) architecture.

Key Projects

• Step-1V Model: In 2024, Stepfun introduced Step-1V, a multimodal large language model (LLM) with over 100 billion settings to improve accuracy.
• Step-2 Model & more: In 2025, the company has already launched the Step-2 model, which exceeded 1 trillion parameters. This model is designed to improve complex thinking and problem-solving abilities.
• Stepfun’s portfolio also includes Step-1.5V, a powerful multimodal large model, Step-1V, an innovative image generation model, and the world’s largest open-source video generation model, Step-Video-T2V, with 30 billion parameters.

🚀 Introduction of Step-Video-TI2V:
a state-of-the-art text-driven image-to-video generation model, capable of generating videos up to 102 frames based on both text and image inputs

💡 Key Features:
– Motion control: stability & flexibility
– Multiple camera movements
– Excels… pic.twitter.com/NxiOWnYFDN

— StepFun (@StepFun_ai) March 25, 2025

Komninos Chatzipapas, founder of AI girlfriend app HeraHaven.AI, told Techopedia:

“Stepfun is funded by investors, including Tencent, but there aren’t any exact figures I could find. But it’s speculated to be in the hundreds of millions of dollars.”

2. Zhipu AI

Zhipu AI is a Chinese artificial intelligence company founded in 2019 by Tang Jie and Li Juanzi. Based in Beijing, Zhipu AI develops large language models and AI technologies.

Key Projects

• GLM Series: Zhipu AI has developed the general language model (GLM) series, including GLM-4, a large language model that can understand long conversations better and work with different types of data, such as text and images.
• ChatGLM: In collaboration with Tsinghua University’s Knowledge Engineering Group, Zhipu AI released ChatGLM, a series of AI chat models that are already trained to have smooth and efficient conversations. They also released AutoGLM-Web, a browser extension powered by AutoGLM, which can perform tasks on your browser with simple commands.
• Ying: Launched in July 2024, Ying is a text-to-video model that can generate six-second video clips from text and image prompts within approximately 30 seconds.
• In April 2025, Zhipu AI rebranded to Z.ai and launched the fully open-source GLM-4-0414 model family under the MIT License, including six models across 9B and 32B sizes, with flagship models like GLM-4-32B for general-purpose tasks, GLM-Z1-32B for enhanced reasoning, and GLM-Z1-Rumination-32B for deep research capabilities.

🚀 New name, same mission — ChatGLM is now https://t.co/IQMnfBczAT.

To mark this new chapter, we’re launching the fully open-source GLM-4-0414 model family under the MIT License — use it, build on it, profit from it.

We’re open-sourcing six models across 9B and 32B sizes. Here… pic.twitter.com/eyDoOXJG1F

— Z.ai (@Zai_org) April 15, 2025

3. MiniMax

MiniMax is a Chinese artificial intelligence company founded in December 2021 by former employees of SenseTime, including Yan Junjie, Yang Bin, and Zhou Yucong.

Key Projects

• Talkie: An AI character role-playing app that lets users participate in conversations with virtual characters, including celebrities and fictional characters.
• Hailuo AI: A multimodal LLM consumer platform that offers AI-generated music and text features.
• Video-01: A text-to-video model under Hailuo AI that generates short videos based on textual prompts.
• MiniMax-01 Series: In January 2025, MiniMax released the MiniMax-01 series, including MiniMax-Text-01 and MiniMax-VL-01, introducing the innovative Lightning Attention mechanism. This series supports a 4M-token context window – 20 to 32 times larger than competitors like GPT-4o and Claude-3.5-Sonnet – enabling advanced long-context processing.

This is wild!

Hailuo AI just dropped MiniMax-01, an industry-first foundation models with Lightning Attention.

And it supports mind blowing 4 MILLION token context size, that’s the entire Harry Potter series three times over!

Why it’s a game-changer and how to try: pic.twitter.com/63z0bapTNL

— Min Choi (@minchoi) January 14, 2025

Jason Snyder, inventor and co-founder at SuperTruth, told Techopedia:

“That’s a rare example of Chinese AI finding traction outside its home market. It should be noted that AI companionship is a double-edged sword. While viral adoption is excellent, keeping users engaged long-term is notoriously tricky.”

If Talkie doesn’t evolve beyond novelty, “users will churn the second the next shiny AI thing drops,” according to Snyder. He added:

“MiniMax has the engagement, but can it build the addiction loop? If they figure that out, they could be the TikTok of AI. Otherwise, they’re another Clubhouse — hot today, forgotten tomorrow.”

4. Moonshot AI

Beijing-based Moonshot AI is a Chinese artificial intelligence company founded in March 2023 by Yang Zhilin, Zhou Xinyu, and Wu Yuxin.

Key Projects

One of Moonshot AI’s key projects is Kimi, an AI chatbot launched in October 2023. Kimi stands out because it can handle massive amounts of text – up to 2 million Chinese characters – in a single input.

In November 2024, Moonshot AI launched K0-math, a new Chinese AI model built for solving math problems. K0-math is designed to solve tough math problems by breaking them down into smaller steps. If it makes a mistake, it can try again to find the right answer.

In January 2025, Moonshot AI launched Kimi 1.5, an enhanced version of its chatbot. The company says Kimi 1.5 is as good as OpenAI’s o1 in certain skills, such as math, coding, and understanding various types of information.

Kimi k1.5: The Multimodal Reasoning Model
– Available now on https://t.co/7YeIld97BT 🦄

💡 What can Kimi k1.5 do?

🔹 Image to Code: Convert images into structured code and insights
🔹 GeoGuessr: Identify and pinpoint locations in geography games like a pro 🌍
🔹 Visual… pic.twitter.com/LEnTC4zZRL

— Kimi.ai (@Kimi_Moonshot) January 26, 2025

Snyder told Techopedia:

“Kimi 1.5 has demonstrated superior performance in coding, mathematics, and logical reasoning, achieving state-of-the-art results across multiple benchmarks and modalities. If AI is an IQ test, Moonshot is scoring off the charts. But the best AI doesn’t always win.”

OpenAI isn’t the leader just because of GPT; it’s because of distribution. Moonshot doesn’t have that yet. Snyder added:

“Pure brains won’t cut it – [Moonshot] needs a killer strategy,” Snyder said. If they solve distribution, they could be China’s answer to OpenAI. If not, they’ll be a footnote in AI history.”

However, the company has no intention to stop. In April 2025, Kimi released Kimi-VL and Kimi-VL-Thinking, the latest open-source lightweight yet powerful Vision-Language Models with reasoning capabilities. Later, in the same month, Moonshot introduced Kimi-Audio, an open-source audio foundation model with advanced capabilities in audio understanding, generation, and conversation.

5. Doubao

Doubao is a Chinese artificial intelligence initiative by ByteDance, the company behind TikTok. Launched in August 2023, Doubao has quickly become China’s most popular AI chatbot.

Key Projects

• PixelDance, launched in October 2024, is an AI video generation tool that creates smooth, high-quality animations and short videos.
• Seaweed, also released in October 2024, is another AI video generation tool that creates realistic video content from text descriptions, helping content creators and marketers.
• Dreamina (Jimeng), launched in late 2024, is an AI-powered tool that lets users create high-quality images and videos just by typing descriptions or uploading pictures. It also includes smart editing features.
• Doubao-1.5 Pro, released in January 2025, is an advanced AI model that aims to improve reasoning and problem-solving skills.

Doubao 1.5 pro is launched! Our effective MOE models achieve SOTA-level performance on llm & multimodal benchmarks. What’s more, we have open sourced the LLM RL library used for Doubao llm/multimodal RL(HF) and O1-like reasoning, available at https://t.co/ao3YtSGMWb.

The team… pic.twitter.com/OCpj9aZnVY

— Haibin (@eric_haibin_lin) January 22, 2025

Pam Aungst Cronin, founder at Pam Ann AI, told Techopedia:

“Bytedance’s Doubao could be considered the ‘ChatGPT of China,’ with its high level of adoption (now up to 17 million daily active users). Bytedance is also prioritizing the development of Dreamina, a video tool that competes with OpenAI’s Sora.”

The Bottom Line

According to Snyder, the AI battle isn’t just about better models; it’s about who controls the data.

“China is racing to build better AI and control the data that feeds it. AI doesn’t innovate in a vacuum; it evolves based on the information provided. When that data is tightly controlled, AI stops being a tool for discovery and becomes a machine for reinforcing narratives,” Snyder said.

“The US must take the lead, not just in building AI, but in ensuring it remains open, driven by the free flow of data and ideas. If China sets the global standard, we risk losing not just technological dominance but control over the narratives shaping the future.”

FAQs