DeepSeek-V4 Technical Report Author List Released Amid Core Team Talent Drain

On April 28, Daily Economic News reported that DeepSeek recently published a comprehensive 58-page technical whitepaper for its V4 architecture. The accompanying author roster, which credits nearly 300 research and engineering professionals, has rapidly become a focal point for industry observers.

Notably, founder Liang Wenfeng is listed alongside the entire R&D and engineering team rather than holding a separate executive credit. Within this roster, ten contributors are explicitly tagged as "former employees," including prominent technical leads such as Wang Bingxuan, Guo Daya, and Wei Haoran. Public disclosures confirm that since the second half of 2025, at least five core development personnel have officially exited the organization.

Industry tracking reveals that these departing specialists have been absorbed by major technology enterprises: Wang Bingxuan, a lead architect for the initial large language model iteration, has transitioned to Tencent; Luo Fuli, a key contributor to the V2 framework, was recruited by Xiaomi; Guo Daya, a principal researcher on the R1 reasoning model, joined ByteDance's Seed division; Ruan Chong, a multimodal technology expert, moved to autonomous driving firm DeepRoute.ai; and Wei Haoran, a lead author for the OCR pipeline, has departed with his next career move currently unannounced. Collectively, these engineers span four foundational AI domains: base model training, logical reasoning, optical character recognition, and multimodal fusion.

Originally incubated in 2023 by quantitative trading firm High-Flyer Quant, DeepSeek leveraged the open-source release and superior performance of models like DeepSeek-V3 and DeepSeek-R1 to rapidly ascend as a global benchmark in artificial intelligence innovation throughout 2025.