In China, the AI company DeepSeek has introduced its highly anticipated V4 large language model (LLM) in a preview release. This significant launch comes more than a year after the company's previous model, R1, which had garnered attention for its cost-effectiveness and performance, leading to controversy around accusations of intellectual theft in Silicon Valley. The V4 series features two versions, a 'Pro' model with 49 billion activated parameters and a 'Flash' model with 13 billion activated parameters, both capable of processing a context length of up to 1 million tokens. The V4-Pro-Max configuration is said to redefine the standards for open models, improving upon previous versions by significantly narrowing the performance gap with leading commercial models, including Google's Gemini 3.1-Pro and OpenAI's GPT-5.2. Notably, in core tasks, DeepSeek claims that the V4 model outperforms prior iterations and competes closely with top models in terms of knowledge evaluation and reasoning benchmarks—showing solid performance against industry giants like Anthropic’s Claude Sonnet 4.5. Huawei has pledged support for the V4 series via its advanced Ascend AI chips, showcasing the growing collaboration between leading tech firms in China. Meanwhile, speculation regarding DeepSeek's potential first funding round has intensified, with notable Chinese tech entities like Tencent and Alibaba reportedly in negotiations to participate, indicating strong investor interest. Sources suggest the company could be valued at around $40 billion, with Tencent proposing to acquire a significant stake in the venture. Amid this landscape, DeepSeek's Chinese competitors have also ramped up their efforts, unveiling their own AI models to keep pace with the anticipated impact of the V4. New releases from competitors include Alibaba's Qwen3.5 and ByteDance's Seedance 2.0, all vying for leadership in the rapid evolution of AI technologies. As the global AI race continues, DeepSeek's V4 model represents a pivotal moment for the company and the broader field of open-source AI.