Ovis2 is the latest version of the Ovis series models proposed by Alibaba's international team. Compared to the previous version 1.6, Ovis2 has significant improvements in data construction and training methods. It not only enhances the capacity density of small models but also greatly improves chain of thought (CoT) reasoning capabilities through instruction fine-tuning and preference learning. Additionally, Ovis2 introduces video and multi-image processing capabilities, and enhances multilingual abilities and OCR capabilities in complex scenarios, significantly increasing the model's practicality.

Recently, DeepSeek announced a significant update on its official X account, stating that it will progressively open source five code repositories over the next week. The company reported that the currently completed online service module has undergone rigorous testing and deployment, making it ready for production use. DeepSeek emphasized that the intention behind sharing the code is to ensure that every line of code becomes a powerful driving force for advancing the industry. The company stated that there is no so-called ivory tower in this field, but rather a pure spirit of garage entrepreneurship and community building.

On February 21, 2025, Beijing time, the DeepSeek AI team, focused on exploring Artificial General Intelligence (AGI), made an exciting announcement on social media, officially kicking off their participation in the #OpenSourceWeek event. This small yet ambitious tech team announced that starting next week, they will open source five code repositories to share their research progress transparently with the global developer community.

Recently, Baidu founder Li Yanhong elaborated on the reasons behind the company's decision to open source the Wenxin large model 4.5 series during Baidu's Q4 2024 and full-year earnings call. Li emphasized that Baidu gained valuable insights from the DeepSeek project, namely that open sourcing the best models can significantly promote user adoption and recognition. He explained that when a model is open sourced, people's curiosity drives them to experiment, and this natural exploratory behavior helps the model be more widely applied in real-world scenarios.

Recently, xAI, the artificial intelligence company founded by Musk, officially launched its latest chatbot – Grok 3. According to the company's official statement, Grok 3 has performed exceptionally well in various tests, even competing with industry giants like Gemini, DeepSeek, and ChatGPT. This news undoubtedly brings new expectations and discussions to the AI field. Grok 3 is currently available to Premium + subscribers, meaning users can experience the powerful features of this new model first.

On February 18, Kunlun Wanwei announced the open source of China's first AI short drama creation video generation model SkyReels-V1 and China's first SOTA-level expression action controllable algorithm based on video base model SkyReels-A1. This initiative aims to address current issues such as the lack of open-source AI video generation models, high costs, and usability challenges, promoting innovation and development in the AI short drama industry. SkyReels-V1 is China's first open-source video generation model for AI short drama creation. The model combines lip-sync generation.

In the field of artificial intelligence, the rise of DeepSeek has sparked widespread attention and concern in the West. However, Mistral CEO Arthur Mensch has welcomed this emerging force, believing that DeepSeek's new models will promote the development of the open source ecosystem. Mistral, dubbed the 'European OpenAI,' has consistently advocated for open source principles, emphasizing the importance of openness and collaboration since its establishment in 2023. In an interview with Business Insider,

Today, Baidu announced an important plan: in the coming months, it will gradually launch the new Ernie Large Model 4.5 series, which is scheduled to be officially open-sourced to the public on June 30. This initiative marks Baidu's continued dedication and innovation in the field of artificial intelligence.

Recently, Huggingface, the world's largest open source AI community, released the latest Open LLM Leaderboard. The results show that all ten of the top-ranked open source large models are derivative models that have been fine-tuned based on Alibaba Tongyi Qwen's (Qwen) open-source model. This achievement marks Qwen's dominant position in the open-source AI field and further enhances its global influence. The Open LLM Leaderboard is widely...

Recently, Tongyi Laboratory officially open-sourced its innovative music generation technology - InspireMusic. This technology aims to create an open-source AIGC toolkit that integrates music, song, and audio generation capabilities, providing researchers, developers, and music enthusiasts with a comprehensive creative platform. InspireMusic offers researchers and developers rich tools for training and fine-tuning music/song/audio generation models, as well as efficient models to optimize generation results. Meanwhile, this toolkit...

Fei-Fei Li, a Stanford University computer scientist and founder of a startup, known as the 'Godmother of AI,' has proposed the 'three fundamental principles for future AI policy making' ahead of the upcoming Paris AI Action Summit next week. She emphasizes that AI policies must be based on 'science rather than science fiction.' Li believes that policymakers should focus on the realities of AI, rather than constructing grand future scenarios built on utopia or apocalyptic visions. She specifically points out that chatbots and co-pilot programs are not 'intelligent agents with intent, free will, or consciousness.'

Recently, PramaLLC launched its latest artificial intelligence model - Background Erase Network BEN2. This model has made significant innovations in foreground segmentation technology. BEN2 uses a Confidence Guided Matting (CGM) pipeline, employing a refined network specifically designed to handle pixels with lower confidence from the base model, resulting in more precise and reliable cut-out effects.