DeepSeek V4, spearheaded by Liang Wenfeng, is anticipated to launch in April 2026, featuring enhanced code capabilities and a focus on long-term memory optimization tailored for China's computing ecosystem. Concurrently, Tencent's HunYuan model, led by Yao Shunyu, is also expected to debut in April, boasting approximately 30 billion parameters with an emphasis on long-context capabilities and Agent task evaluation. The release of these models marks a significant step in the competitive landscape of domestic large model development, particularly in multimodal capabilities and practical application performance.