Chinese AI Startup Zhipu Launches AutoLM Rumination, Challenges DeepSeek R1

Chinese AI startup Zhipu AI has introduced a powerful new AI agent, AutoLM Rumination, capable of conducting in-depth research, generating detailed reports, and assisting users in planning various activities—all at no cost. According to a Reuters report, the AI agent is powered by Zhipu’s proprietary models, which are said to rival DeepSeek R1 in performance. Additionally, the company claims that its model operates eight times faster while requiring only a fraction of the computational resources—one-thirtieth to be precise.

Zhipu AI has been rapidly expanding, recently securing $69 million in Series D funding, following closely after a substantial $137 million funding round. This latest valuation places the company at $2.74 billion. Over the years, Zhipu has received backing from major Chinese tech giants like Alibaba and Tencent, along with financial support from Chinese state-owned entities. However, its global expansion efforts have been hindered by U.S. trade restrictions. In January, the U.S. Department of Commerce added Zhipu and its subsidiaries to its export control entity list, preventing access to American-made components.

The unveiling of AutoLM Rumination arrives amid significant developments in China’s AI landscape. Just weeks ago, another Chinese AI company introduced Manus AI, an advanced general-purpose AI agent that made waves by outperforming OpenAI’s deep research tool on the GAIA benchmark, a key test for AI models solving real-world problems. Unlike Zhipu’s free offering, Manus AI is available through a subscription model, with a standard plan priced at $39 per month and a premium version costing $199 per month.

China’s AI industry is currently witnessing remarkable growth, with tech giants Alibaba and Tencent rapidly rolling out new AI models. Alibaba recently introduced QwQ 32B, an AI model boasting 32 billion parameters, designed to deliver performance on par with DeepSeek R1, which features 671 billion parameters, of which 37 billion are active during processing. QwQ 32B has demonstrated superior results in benchmarks related to coding, mathematical reasoning, and general problem-solving, even surpassing OpenAI’s o1-mini model in several areas.

Meanwhile, DeepSeek AI continues to make strides in AI development. Its DeepSeek-V3 model has been updated, with the latest version, DeepSeek V3-0324, securing the top spot among non-reasoning models in various performance benchmarks. Furthermore, reports indicate that DeepSeek R2 is in the pipeline and could be released sooner than initially planned. The company had aimed for an early May launch but is now considering accelerating the timeline.

With Chinese AI firms aggressively advancing their technology and rolling out competitive models, the AI race is heating up, signaling intensified competition not just within China but on a global scale.