DeepSeek V4 Pro: Enhanced AI Agent Capabilities

Chinese AI company DeepSeek has launched the latest version of its flagship V4 Pro model, DeepSeek-V4-Pro-0813. This new model significantly enhances capabilities in AI agents and software engineering. It's now available through DeepSeek's website, mobile application, and API. This upgrade allows the model to handle more complex tasks, moving beyond simple user queries. AI agents can break down tasks, use external tools, and make decisions with less human intervention. The new model scored 62.7 on the DeepSWE benchmark for software engineering problems, a notable increase from the preview version's 12.8. It also performs well in terminal operations and cybersecurity tasks. DeepSeek is maintaining relatively low API prices for now, at 3 yuan per million input tokens and 6 yuan per million output tokens. However, pricing will adjust from August 17th, with peak-hour output tokens costing 27 yuan per million. This development is important because AI applications are moving towards more complex tasks like coding and task execution, making AI agents crucial for future software development.