DeepSeek V4 and a new Tencent Hunyuan model will launch in April 2026, according to Chinese tech outlet Whale Lab.
The simultaneous releases signal a strategic pivot among China’s leading AI labs toward practical capabilities and real-world deployment over benchmark competition. Both models enter a crowded domestic market where Alibaba, ByteDance, and others have recently shipped updated systems.
DeepSeek V4 functions as a multimodal large model handling text, image, and video generation natively, according to Whale Lab. The model introduces advances in coding capabilities and long-term memory, a persistent technical challenge for large language models. Earlier reports from the Financial Times had indicated a March release, but the timeline has shifted to April. A “V4 Lite” variant briefly appeared on DeepSeek’s website on March 9, suggesting the broader model family nears completion.
Two models labeled “Healer Alpha” and “Hunter Alpha” recently surfaced on the OpenRouter platform, Whale Lab reported. Hunter Alpha describes itself as a trillion-parameter model with a one-million-token context window for agent workflows. Healer Alpha is characterized as a multimodal system with cross-modal reasoning. No official entity has claimed either model.
DeepSeek has partnered with Huawei and Cambricon to optimize V4 for domestic Chinese AI chips, according to reports. The company has also formed a partnership with Baidu to enhance AI search capabilities.
Tencent’s Hunyuan model will be led by Shunyu Yao, a former OpenAI researcher whom Tencent appointed chief AI scientist in December 2025. The model contains approximately 30 billion parameters and focuses on in-context learning and agent usability, according to Caixin. Yao, 28, graduated from Tsinghua University’s Yao Class and contributed to AI agent frameworks including ReAct and Tree of Thoughts. He has advocated for scenario-driven applications over benchmark optimization, Caixin reported.