Tech Behind New Free AI Agent ‘Closely Rivals or Outperforms GPT-4’

執筆者
J.R. Johnivan
J.R. Johnivan
Apr 1, 2025
2 minute read
Stock photo of China's flag.

Image: Envato/yavdat

eWeek のコンテンツおよび製品のおすすめは、編集上の独立性を保っています。パートナーへのリンクをクリックすると、当社が報酬を得る場合があります。 詳細を見る

Zhipu AI, a Chinese AI startup founded in 2019, recently released its free AI agent to the general public. Known as AutoGLM Rumination, the new solution, which has already secured millions of dollars in government-backed funding, is making headlines across the industry.

What is AutoGLM Rumination?

AutoGLM Rumination is one of the newest AI agents to hit the consumer market. It’s capable of performing basic web searches as well as more advanced research tasks. Current uses for AutoGLM Rumination include technical writing and travel planning.

The AI agent is powered by two of Zhipu AI’s proprietary large language models (LLMs): GLM-4-Air-0414 and GLM-Z1-Air. To assess how these LLMs measure up against competing models, it’s essential to examine available benchmark data.

Sizing up the competition through benchmarks

Though only in operation for a few years, Zhipu AI developers have made ambitious claims regarding the performance of their generative AI tools. Developers claim that GLM-Z1-Air performs eight times faster than DeepSeek-R1; it reportedly does so while only using a fraction of the computational power.

A research paper published in June 2024 shows that Zhipu AI’s most recent LLM, GLM-4, does surpass OpenAI’s GPT-4 across numerous benchmarks. The paper’s authors stated that GLM-4 “closely rivals or outperforms GPT-4 in terms of general metrics such as MMLU, GSM8K, MATH, BBH, GPQA, and HumanEval.”

However, it falls short when compared to other types of AI models, such as Claude 2, Claude 3 Opus, Gemini 1.5 Pro, and GPT-4 Turbo, in certain areas. In Python and Java programming, for example, GLM-4 is the only Zhipu AI model that scores high enough to provide any real competition — and even that lags behind the top models on NaturalCodeBench.

Advertisement

While GLM-4 is leading many of the benchmarks on LongBench-Chat, GLM-4-Air and GLM-4-9B-Chat didn’t perform quite as well. They both struggled to keep pace with the competition in the English language benchmarks, but all three performed well with the Chinese language tests.

Contributing to the open-source community

Zhipu AI has made numerous contributions to the open-source AI community, and they’ve accumulated more than 10 million downloads for their past releases. These include:

  • ChatGLM-6B
  • GLM-4-9B
  • GLM-4V-9B
  • WebGLM
  • CodeGeeX

Despite a slew of AI startups entering the market with their own solutions, developers with Zhipu AI are already making their presence known. AutoGLM Rumination is only the latest in a line of AI-driven products, and we’ll likely hear more from Zhipu AI in the near future.

J.R. Johnivan

J.R. Johnivan

Contributing Expert

J.R. Johnivan is a 17-year veteran whose writing is focused on innovation and technology, including IT, computer networking, security, cloud computing, staffing, human resources, real estate, sports, entertainment, and more.

eWeek Logo

eWeek has the latest technology news and analysis, buying guides, and product reviews for IT professionals and technology buyers. The site's focus is on innovative solutions and covering in-depth technical content. eWeek stays on the cutting edge of technology news and IT trends through interviews and expert analysis. Gain insight from top innovators and thought leaders in the fields of IT, business, enterprise software, startups, and more.

TechnologyAdvice が所有・運営しています。 © 2026 TechnologyAdvice. 無断転載を禁じます

広告主に関する開示:このサイトに掲載されている製品の一部は、TechnologyAdvice が報酬を受け取っている企業のものです。この報酬は、製品がこのサイトのどこにどのように表示されるか(表示される順序など)に影響する場合があります。TechnologyAdvice は、市場で入手可能なすべての企業やすべての種類の製品を掲載しているわけではありません。