7 Hrs_1777370074079_638186083.jpg&w=64&q=75)
Source:
تكنولوجيا
تكنولوجيا
Ready to play
Ready to play
An experiment conducted by Vals found that China’s open-source DeepSeek models are approaching OpenAI’s Daybreak models in their ability to perform cyber tasks, while having fewer restrictions and being available for local download and modification. The Daybreak models completed an average of 81% of the tasks, compared with about 73% for DeepSeek. The cost of an attack using the GPT-6 Sol model ranged from $17 to $46, compared with about $3 using a DeepSeek model. The findings warned that the lower cost of open models and the ease of removing their restrictions could enable them to be used to identify vulnerabilities and exploit them in cyberattacks.
Notice: This Is an AI-Generated Summary
Comments (0)