Chinese artificial intelligence firm DeepSeek formally introduced its V4 Pro model on Thursday, seeking to recover ground from rapidly advancing local competitors while boosting recruitment, computing resources and capital raising. The company stated that V4-Pro-0813 significantly improves agent functions and is accessible via API, mobile application and web platforms. It also announced higher API rates for the V4 Pro and V4 Flash versions along with peak and off-peak pricing tiers.
The launch draws attention following the unexpected strong performance of the lower-cost V4 Flash model over the April preview of V4 Pro in multiple independent evaluations. This outcome was notable since the Pro version is intended as the more advanced offering, indicating swift technological progress between the preview and official release.
DeepSeek gained prominence among Chinese AI developers after its R1 model gained widespread attention in early 2025, prompting discussion on whether advanced AI could be developed at substantially reduced expense compared to American counterparts.
Its initial advantage has faced pressure from successive launches by domestic rivals such as Moonshot AI, Zhipu AI, MiniMax, Alibaba and ByteDance.
The company has encountered challenges in converting its visibility into sustainable operations. Reports from July indicated plans for fresh funding at a valuation near $74 billion, shortly after securing roughly $7.4 billion in its initial external round in June.
This financing represented a shift for a firm previously resistant to outside investment and highlights the rising expenses of AI competition, which demands major outlays on processors, data facilities and expert personnel.
DeepSeek aims to at least double its workforce across units, including data-centre and AI-agent groups. It has also expanded recruitment of chip-design specialists to create proprietary AI processors, potentially lowering dependence on suppliers like Nvidia and Huawei.
