The Chinese AI ecosystem is heating up—fast. In a bold move that could reshape the competitive landscape of AI agents, Beijing-based startup Zhipu AI has launched a new AI agent called AutoLM Rumination, claiming performance on par with DeepSeek-R1—while operating eight times faster and using just one-thirtieth of the compute resources.
Even more disruptive? It’s completely free.
According to Reuters, Zhipu AI unveiled AutoLM Rumination on Monday, describing it as a “deep research assistant” capable of performing complex reasoning tasks, generating highly detailed reports, and helping users with planning and organization across a wide range of domains.
An AI Agent Built for Speed and Efficiency
AutoLM Rumination is powered by Zhipu’s proprietary large language models, which the company asserts match the performance of DeepSeek-R1—China’s current poster child for cutting-edge AI. However, Zhipu says its model is eight times faster and requires a fraction of the computing power (about 1/30th), an impressive feat in a field where resource optimization is becoming just as critical as raw performance.
This combination of speed, efficiency, and cost-effectiveness could make AutoLM Rumination an attractive alternative for developers, researchers, and businesses looking for advanced AI capabilities without the high price tag.
Massive Momentum: $206M in Funding in Weeks
Zhipu AI’s momentum is as noteworthy as its technology. The startup recently raised $69 million in a Series D round, shortly after bagging $137 million in a previous round—bringing its total valuation to $2.74 billion.
Zhipu has attracted investment from some of China’s biggest tech giants, including Tencent and Alibaba, as well as support from state-owned enterprises, reflecting its strategic importance in China’s race for AI dominance.
However, it hasn’t all been smooth sailing. In January, the U.S. Commerce Department placed Zhipu and its subsidiaries on the export control entity list, effectively restricting access to U.S.-made chips and components—adding yet another layer of complexity to China’s AI ambitions.
China’s AI Agent Arms Race
Zhipu’s launch comes just weeks after the introduction of Manus AI, another general-purpose agent developed by a Chinese company. Manus AI surprised many by outperforming OpenAI’s research agent in the GAIA benchmark, a rigorous test of real-world problem-solving. While Manus AI comes with a monthly subscription fee starting at $39, Zhipu’s AutoLM Rumination is being offered for free—at least for now.
China's AI sector is experiencing a flurry of innovation. In parallel, Alibaba recently introduced QwQ 32B, a 32-billion-parameter model that reportedly rivals the performance of DeepSeek-R1, which boasts 671 billion parameters (37 billion activated). Interestingly, QwQ 32B outperformed OpenAI's o1-mini in several benchmarks, especially in coding and mathematical reasoning—further signaling a shift in global AI leadership.
DeepSeek: The Benchmark to Beat
DeepSeek has become the gold standard for AI performance in China, particularly after the success of its DeepSeek-V3 update. The latest version, DeepSeek-V3-0324, currently tops the charts in non-reasoning benchmark tests, making it a tough act to follow.
But Zhipu AI may be signaling that the race is no longer just about raw benchmark power—it’s about speed, accessibility, and efficiency. The company’s new AI agent seems designed not only to rival DeepSeek but also to democratize high-end AI tools at a time when paywalled access is becoming the norm.
And the competition isn’t over. DeepSeek R2 is already in the works, with a potential early release date being considered, according to reports.
Final Thoughts: A Shift in AI Dynamics
Zhipu AI’s launch of AutoLM Rumination represents more than just a new product—it’s a strategic play in an escalating AI arms race. By offering powerful AI tooling for free, and at drastically reduced compute costs, Zhipu is challenging the prevailing narrative that top-tier AI is exclusive or expensive.
As the Chinese AI ecosystem continues its rapid ascent, global players will be watching closely. The race is no longer just about innovation—it’s about accessibility, efficiency, and who can bring the next generation of AI tools to the masses first.