Liang Wenfeng Reveals Internally: DeepSeek V4 Will Be Released in Late April

Leak and schedule

It has been reported that Liang Wenfeng (梁文峰), a senior figure at Chinese AI firm DeepSeek, told staff in an internal briefing that DeepSeek V4 is slated for a late‑April release. The remark, shared to company channels and later picked up by ifeng, was described as an internal roadmap update rather than a formal public launch announcement; the company has not publicly confirmed the date. Reportedly, the V4 rollout is meant to follow a period of closed testing and incremental safety hardening.

Industry backdrop: hype, safety and narrative positioning

China’s fast‑moving AI sector is now moving in lockstep with global narratives about model capability and risk. It has been reported that this week’s debate over Anthropic’s “Mythos” marketing — with prominent voices such as Yang Likun (杨立昆) and NYU’s Gary Marcus calling the claims overblown — has highlighted how vendors blend technical claims with safety rhetoric to shape public perception. Why does this matter? Because model upgrades are as much marketing events as technical milestones; firms routinely couple product pushes with safety pledges to occupy regulatory and public attention.

Geopolitics, regulation and what to watch

DeepSeek’s planned V4 launch comes at a sensitive moment. U.S. export controls on advanced chips, sanctions and heightened scrutiny of AI supply chains are forcing Chinese developers to accelerate software advances while managing access to foreign hardware. Will V4 be a genuine step‑change or another carefully staged narrative? Observers will watch release notes, safety guardrails and third‑party audits closely — and it has been reported that DeepSeek has so far declined to provide further comment to media inquiries.