DeepSeek has released an updated version of its V4 Pro large language model, the 0813 build. Benchmarks indicate it slightly lags behind Anthropic's Claude Fable 5, but DeepSeek V4 Pro is significantly more affordable, costing approximately 4,600% less. Meanwhile, Anthropic's new AI output watermarking system, intended to meet EU AI Act requirements, faces user backlash. Some users express concerns that the watermarking could reveal their use of the chatbot for professional and academic purposes, while others view it as a crucial step for transparency.
DeepSeek has introduced an updated iteration of its V4 Pro large language model, specifically the 0813 build. Comparative benchmarks suggest that this new version offers marginal performance improvements, falling slightly short of Anthropic's Claude Fable 5. However, the cost-effectiveness of DeepSeek V4 Pro is a major distinguishing factor, as it is priced approximately 4,600% lower than Claude Fable 5. This significant price difference makes DeepSeek a more accessible option for users prioritizing budget.
In parallel, Anthropic's implementation of a new AI output watermarking system has ignited controversy among its user base. This system is designed to align with the stipulations of the EU AI Act. A segment of users has voiced strong opposition, citing potential negative repercussions for their professional and academic work. They fear that the watermarking could inadvertently expose their reliance on the chatbot for tasks, potentially impacting their employment or academic standing. Conversely, another group of users supports the watermarking initiative, viewing it as an essential measure for enhancing transparency in AI-generated content and ensuring accountability.
The EU AI Act mandates certain transparency requirements for AI systems, including measures to identify AI-generated content. Anthropic's watermarking system is a response to these regulatory demands, aiming to provide a method for distinguishing AI-produced text from human-written content. The debate highlights a broader tension between the drive for AI innovation and the need for regulatory oversight and user privacy.
DeepSeek has introduced an updated iteration of its V4 Pro large language model, specifically the 0813 build. Comparative benchmarks suggest that this new version offers marginal performance improvements, falling slightly short of Anthropic's Claude Fable 5. However, the cost-effectiveness of DeepSeek V4 Pro is a major distinguishing factor, as it is priced approximately 4,600% lower than Claude Fable 5. This significant price difference makes DeepSeek a more accessible option for users prioritizing budget.