DeepSeek officially launched V4-Pro-0813, its flagship model now available via app, web, and API. Th...
The AMW Read
DeepSeek's V4 Pro launch and pricing shift update its player map position, while the $74B valuation explicitly triggers capital-cycle tagging.
DeepSeek officially launched V4-Pro-0813, its flagship model now available via app, web, and API. The company said the new version is significantly improved at supporting AI agents, a growing area of focus as developers move from question-answering to multi-step task execution. Notably, the cheaper V4 Flash model outperformed the V4 Pro preview in several independent tests, underscoring the rapid pace of DeepSeek's model development.
Alongside the launch, DeepSeek announced major API pricing changes effective August 17, with separate peak and off-peak rates for V4 Pro and V4 Flash. Price increases range from 50% to over 1,100% depending on model and token type. This marks a decisive shift from DeepSeek's earlier low-cost strategy, which had pressured US and Chinese rivals. The move reflects a maturing commercial approach as the company faces intensifying domestic competition from Moonshot AI, Zhipu AI, MiniMax, Alibaba, and ByteDance.
DeepSeek is also reportedly seeking to raise nearly $8 billion in a new funding round that would value the company at about $74 billion, following an earlier $7.4 billion external round. The valuation signals strong investor confidence in DeepSeek's commercialization prospects despite rising training and operating costs. For builders, the API price hike suggests DeepSeek is prioritizing revenue sustainability over market share, and reliance on its cheapest inference may become more expensive. Investors should watch whether DeepSeek can maintain performance leadership while absorbing higher infrastructure costs.
