DeepSeek LLM 67B Large Model Released for Free Commercial Use Without Application


The strength of open-source AI large models in China is highlighted, with DeepSeek, Qwen, and Kimi tying for first place in the latest ranking. The list was published by AI researchers, covering 35 institutions, with more than half being Chinese teams, demonstrating the rapid development of China's open-source field, contrasting with the U.S. closed-source model.
OpenAI released the GPT-5.2 series model, positioned for daily professional use, aiming to enhance user economic value. The series includes the Instant, Thinking, and Pro versions, with significant improvements over GPT-5.1 in tasks such as spreadsheets, presentations, code writing, long text understanding, and image processing.
Xiaomi elevates AI large models as its core strategy for the next decade, with quarterly investment growth exceeding 50% over the past year. The company has launched a global talent recruitment drive, offering salaries up to 10 million yuan per position to address talent shortages. Former DeepSeek core members have joined, unveiling the MiMo team.....
Xiaomi executive Lu Weibing announced increased AI investment, with significant progress in large models and applications exceeding expectations. He emphasized integrating AI with reality as the future direction and is actively recruiting talent.....
DeepSeek releases V3.2 Standard and Deep Thinking editions. V3.2 matches GPT-5 in 128k context, while Deep Thinking rivals Gemini 3 Pro in benchmarks and achieves gold medal scores in IMO 2025. Key upgrade: sparse attention reduces computational load and memory usage, enabling million-token single-GPU inference for the first time in open-source models.....