OpenAI GPT-4 Ranks Highest for Lowest Hallucination Rate in Document Summarization, According to Vectara

In the Vectara rankings, OpenAI GPT-4 leads the pack with an impressive 97% accuracy rate and a mere 3% hallucination rate. Vectara offers a method for evaluating the hallucination rates of large language models through open-source models, providing businesses with more reliable generative AI systems. Google's model has performed averagely in the rankings, sparking discussions on the development of artificial intelligence and the stability of the technology.

Unveiling the Mystery of MiniMax M2: Why Choose Full Attention Mechanism?

The MiniMax M2 model uses a full attention mechanism, abandoning linear or sparse attention techniques. The development team believes that although the latter can save computing resources, full attention is more efficient in industrial applications and can improve model performance. This decision aims to optimize actual deployment results and promote the development of AI technology.

China Huadian Launches 'Huadian Zhi' Large Model, Energy Management Enters a New Intelligent Era

China Huadian launched the 'Huadian Zhi' large model at the 2025 New Power System Forum, achieving breakthroughs in artificial intelligence and predictive applications. The model pioneered runoff prediction technology, increasing the water energy utilization rate of the Wujiang River Basin from 5.8% to 10.8%, promoting the intelligent transformation of the power industry.

Musk's Prediction for the Future: AI Will Replace Phones and Applications!

The CEO of Tesla, Musk, predicts that artificial intelligence will dominate human life in the next 5-6 years, replacing phones and applications. He believes that devices will become edge nodes for AI interfaces, and people will access services through voice interaction. Technology will completely change the way people live.

Perplexity and Getty Images Reach Long-Term Licensing Agreement to Ensure Content Compliance and Attribution

Artificial intelligence company Perplexity has reached a long-term image licensing agreement with Getty Images, allowing the use of Getty's images in its AI search tool. This marks a significant strategic shift for Perplexity from past controversies over content scraping and plagiarism (such as being accused of stealing Getty images) towards establishing formal and legal content partnerships.

OpenAI Launches Aardvark: An Intelligent Security Research Assistant to Enhance Software Protection

OpenAI has launched Aardvark, an intelligent security assistant based on GPT-5, to help developers and security teams efficiently address the challenge of thousands of new vulnerabilities each year. The tool continuously analyzes source code, automatically identifies vulnerabilities, assesses risks, prioritizes them, and provides remediation solutions, significantly improving the efficiency of software security protection.