DeepSeek has confirmed that its V4 model will drop next week, marking the next major release from the Chinese AI lab that shook up the industry earlier this year with its R1 reasoning model.
Details on V4's architecture and capabilities remain thin, but the timing matters. DeepSeek V3, released in late 2024, proved that a relatively small team could train a frontier-class model for a fraction of what US labs were spending. R1 followed in January 2025 and matched OpenAI's o1 on several reasoning benchmarks while being fully open-weight (meaning anyone can download and run the model locally). Both releases sent ripples through the AI industry and contributed to a brief selloff in AI-related stocks.
The V4 announcement comes during a period of rapid model releases across the industry. OpenAI, Google, and Anthropic have all shipped major updates in recent months, and the performance gap between open-weight and closed models continues to narrow. If V4 follows DeepSeek's pattern of publishing training costs and releasing weights publicly, it will add more pressure to that trend.
No pricing, benchmark numbers, or parameter counts have been shared yet. We will cover the full release details once V4 ships.