Loading timeline…
20262020s
DeepSeek V4
DeepSeek launched V4 in preview on April 24, 2026: an open-weight, roughly 1-trillion-parameter MoE model with a 1M-token context, reaching general availability in July.
Why It Mattered
DeepSeek V4 extended the company's pattern of open-weight releases that pressure closed frontier labs, packing about 1 trillion parameters with only ~37B active per token, a 1-million-token context window powered by Engram conditional memory, and native multimodal generation. Released as V4-Pro and V4-Flash on Hugging Face and its own API, it delivered near-frontier coding performance at a fraction of typical costs, reinforcing China's position in open-weight AI and continuing the downward pressure on global inference pricing.
Who Built It
DeepSeek
Applications
- Open-source AI
- Coding
- Multimodal Generation