Today, the Alibaba Cloud Tongyi team officially released the new mathematical reasoning process reward model Qwen2.5-Math-PRM. This model offers two sizes, 72B and 7B, with performance significantly outperforming similar open-source process reward models, especially excelling in identifying reasoning errors. The 7B version of Qwen2.5-Math-PRM astonishingly surpasses the widely popular GPT-4o, marking an important step in Alibaba Cloud's research and development of reasoning models.

On January 7, 2025, Beijing Zero One Everything Technology Co., Ltd. refuted recent online rumors regarding 'Alibaba acquiring Zero One Everything'. The company clearly stated that these rumors are untrue and denied the related false statements. Zero One Everything emphasized that its collaboration with Alibaba Cloud is based on deep cooperation in technology, computing power, business, and talent, rather than an acquisition relationship.

The latest collaboration between Aliyun and Hezhima Intelligent has sparked a new wave in the domestic smart car sector. Aliyun announced that its Tongyi Qianwen models with 1.5 billion and 3 billion parameters have been successfully ported to the Hezhima Intelligent Wudang C1200 series automotive-grade chip, marking a significant breakthrough for domestic large models in vehicular scenarios. In practical applications, this integrated solution supports multi-turn natural dialogue in offline environments, providing users with a smooth human-computer interaction experience. This technological breakthrough not only demonstrates the R&D capabilities of domestic AI chips but also injects new momentum into the development of smart cars.

The Aliyun Bailing platform recently announced the launch of the Qwen2.5-Turbo million long text model, developed by the Tongyi Qianwen team. The Qwen2.5-Turbo model supports processing ultra-long contexts of up to 1 million tokens, equivalent to 1 million English words or 1.5 million Chinese characters.

In the ever-changing tech landscape, Aliyun has once again made waves with its bold pricing strategy. At the 2024 Yunqi Conference, Alibaba Group not only showcased the future vision of AI technology but also announced a remarkable decision: a price reduction of up to 85% on AI large models. This move will undoubtedly stir the tech community and bring new possibilities for the popularization and application of AI technology. Alibaba Group CEO Wu Yongming delivered an important speech at the conference, emphasizing that the impact of generative AI will far exceed that of mobile applications, profoundly changing our lives.

On September 2nd, Tongyi Qwen announced the open sourcing of its second-generation visual language model Qwen2-VL, and launched APIs for the 2B and 7B sizes as well as their quantized versions on the Aliyun Bailian platform for direct user access. The Qwen2-VL model achieves comprehensive performance improvements in several areas. It can understand images of different resolutions and aspect ratios, setting global leading performance on benchmarks such as DocVQA, RealWorldQA, and MTVQA.

The Institute of Geochemistry of the Chinese Academy of Sciences, in collaboration with Aliyun, launched the world's first "Lunar Science Multi-modal Professional Large Model" at the 2024 Digital Expo. This model is based on Aliyun's Tongyi series and utilizes RAG retrieval-augmented technology, fine-tuned and trained on Aliyun's Baolian exclusive version, specifically aimed at improving the accuracy of distinguishing the ages and shapes of lunar impact craters, currently exceeding 80%. This innovative model combines visual, multi-modal, and natural language processing technologies, and its optimal application is in identifying impact craters on the moon, which is significant for studying lunar geological evolution.

Aliyun has launched the Qwen2-Math series of large language models, focusing on mathematical problem-solving abilities that surpass existing open-source models and rival well-known closed-source models, including GPT-4, Claude-3.5-Sonnet, Gemini-1.5-Pro, and Llama-3.1-405B. This series is based on the Qwen2 model, which has undergone deep pre-training on a specialized mathematical corpus that includes mathematical web texts, professional books, code examples, and exam questions, as well as data autonomously generated by Qwen2.

The ComfyUI of Aliyun PAI Artlab has introduced a workflow for creating Olympic highlight moment posters, allowing users to generate personalized Olympic-themed posters in just three steps. PAI-ArtLab is a one-stop AIGC design platform that offers dataset management, model training, and AI drawing functions, with over 200 mainstream AI generation models built-in. Users can upload local datasets for labeling or training, invoke preset AI models, or use tools such as WebUI, ComfyUI, and Khoya.

Recently, the Aliyun Tongyi Qianwen APP launched a free feature called 'Tongyi Dance King'. Users only need to input a specified password and upload a photo to generate a personalized dance video. The system retains original image features, including facial expressions, body proportions, clothing, and more. It provides 12 popular dance templates, such as Subject Three, Mongolian Dance, and Paddle Dance. This is powered by the self-developed video generation model 'Animate Anyone' from the Aliyun Tongyi Laboratory.

Aliyun has launched the free open-source database AI algorithm PilotScope. PilotScope enables one-click deployment of AI algorithms in databases, improving the speed of tasks such as query optimization by 1 to 2 times. PilotScope has been piloted internally at Aliyun and is now freely available as open-source.

Aliyun plans to expand its business scope to include AI software development. Recently, Aliyun Computing Co., Ltd. made a business change to extend its scope to the development of artificial intelligence application software. This initiative aims to provide developers with more opportunities to promote business growth in the AI era.

Qwen-VL is a large-scale vision language model launched by Aliyun that supports image and text input. Compared to other vision language models, Qwen-VL has added capabilities such as visual localization and image text understanding. Qwen-VL has received over 3,400 stars on GitHub and has been downloaded more than 400,000 times.