Podcast Episode
The simultaneous releases signal a new phase in the country's AI arms race, with both companies targeting practical, real-world capabilities rather than raw benchmark performance.
Intriguingly, two anonymous models recently appeared on the OpenRouter platform under the names Hunter Alpha and Healer Alpha. Hunter Alpha claims to be a trillion-parameter model with a one-million-token context window designed for agent workflows, while Healer Alpha is billed as a multimodal model with cross-modal reasoning. No official entity has claimed either model, but speculation has linked them to DeepSeek V4.
Reports also indicate DeepSeek has partnered with Huawei and Cambricon to optimise V4 for domestic Chinese AI chips, and formed a collaboration with Baidu to strengthen AI search.
Yao, a graduate of Tsinghua University and Princeton, is known for foundational AI agent frameworks including ReAct and Tree of Thoughts. His philosophy of scenario-driven development over benchmarking appears to be shaping Tencent's new strategy.
DeepSeek V4 and Tencent Hunyuan Set for Simultaneous April Launch
March 14, 2026
0:00
2:12
China's AI race heats up as DeepSeek V4 and Tencent's new Hunyuan model both target April 2026 launches. DeepSeek V4 promises multimodal capabilities and a breakthrough in long-term memory, while Tencent's model prioritises real-world agent performance over benchmark scores.
Two Giants, One Month
China's artificial intelligence landscape is about to get a major shake-up. DeepSeek V4 and a new Tencent Hunyuan model are both expected to launch in April 2026, according to a report from Chinese tech outlet Whale Lab published on March 14.The simultaneous releases signal a new phase in the country's AI arms race, with both companies targeting practical, real-world capabilities rather than raw benchmark performance.
DeepSeek V4: Multimodal and Memory-First
DeepSeek V4, led by founder Liang Wenfeng, is described as a multimodal large model capable of handling text, image, and video generation natively. The model is said to deliver a significant leap in coding performance and a breakthrough in long-term memory, a persistent challenge for large language models.Intriguingly, two anonymous models recently appeared on the OpenRouter platform under the names Hunter Alpha and Healer Alpha. Hunter Alpha claims to be a trillion-parameter model with a one-million-token context window designed for agent workflows, while Healer Alpha is billed as a multimodal model with cross-modal reasoning. No official entity has claimed either model, but speculation has linked them to DeepSeek V4.
Reports also indicate DeepSeek has partnered with Huawei and Cambricon to optimise V4 for domestic Chinese AI chips, and formed a collaboration with Baidu to strengthen AI search.
Tencent Bets on a New Leader
Tencent's new Hunyuan model will be the first major release under Shunyu Yao, a former OpenAI researcher appointed as Tencent's chief AI scientist in December 2025. At roughly thirty billion parameters, the model focuses on in-context learning and agent usability rather than chasing leaderboard rankings.Yao, a graduate of Tsinghua University and Princeton, is known for foundational AI agent frameworks including ReAct and Tree of Thoughts. His philosophy of scenario-driven development over benchmarking appears to be shaping Tencent's new strategy.
A Crowded Field
The dual launches arrive amid a broader wave of Chinese AI releases from Alibaba, ByteDance, and others, intensifying competition across multimodal capabilities, long-term memory, and agent performance.Published March 14, 2026 at 11:11pm