Baidu recently launched its latest Ernie 4.5 and Ernie X1 large models, both available for free trial on the Ernie Bot official website. Ernie 4.5, Baidu's first native multimodal large model, excels in multimodal understanding and logical reasoning, outperforming GPT-4.5 in various benchmark tests. Its API price is only 1% of GPT-4.5's, attracting significant attention from developers and businesses. Ernie 4.5 demonstrates remarkable advancements in multimodal understanding, showcasing...

Recently, Professor Zhai Jidong's team from the Institute of High Performance Computing at Tsinghua University, in collaboration with Tsinghua-affiliated innovative enterprise, Qingcheng Extreme Intelligence, announced the open-sourcing of Chitu, a groundbreaking large model inference engine. This innovative technology marks another significant breakthrough for China in the AI field, particularly in inference engine development. A core highlight of the Chitu engine is its ability to natively run FP8 precision models on non-Nvidia Hopper architecture GPUs and various domestic chips.

Against the backdrop of the continuously expanding artificial intelligence field, Beijing Zhipu AI Technology Co., Ltd. (hereinafter referred to as "Zhipu") recently received a 500 million RMB strategic investment from Zhuhai Huafa Group. This financing marks Zhipu's second funding round within a month. Notably, Zhipu secured 1 billion RMB in investment from Hangzhou state-owned assets and other institutions in early March. This series of capital injections will empower Zhipu to drive innovation and ecological development of its core technology – the GLM foundational large model. On March 13th, Zhuhai Huafa Group and Zhipu signed a cooperation agreement, and both parties...

Shanghai Jieyue Xingchen Intelligent Technology Co., Ltd. and Zhiyuan Robotics have officially signed a deep strategic cooperation agreement. Both parties will conduct in-depth cooperation in the fields of base large models and robot research and development, jointly exploring technological breakthroughs and application innovations of "large model + embodied robot." This cooperation involves world model technology research and development, data cooperation in the field of embodied intelligence, and the implementation of application scenarios such as new retail, aiming to promote the large-scale application of embodied intelligence technology in areas such as home services, new retail, and intelligent manufacturing.

Alibaba has announced the launch of its flagship AI application, New Quark. Built upon Alibaba Tongyi's leading reasoning and multimodal large models, New Quark is a comprehensively upgraded, boundless "AI Super Box." New Quark boasts powerful reasoning capabilities and enables multimodal interaction, allowing users to interact with AI in real-time. The AI will think and act based on user needs, continuously adjusting strategies to ensure task completion. This innovative design provides users with both convenience and an engaging experience.

On March 13th, Alibaba officially launched its AI flagship application—New Quark. This newly upgraded Quark, built upon Alibaba Tongyi's leading reasoning and multimodal large models, creates a boundless AI Super Box, offering users a brand-new AI experience.

Google CEO Sundar Pichai announced at a launch event that Google has open-sourced its latest multimodal large model, Gemma-3. This model is attracting significant attention for its low cost and high performance. Gemma-3 offers four different parameter scale options: 1 billion, 4 billion, 12 billion, and 27 billion parameters. Surprisingly, the largest 27 billion parameter model only requires a single H100 GPU for efficient inference, while similar models often require ten times the computing power.

Today, the Doubao Large Model team officially released a technical report on its text-to-image technology, publicly disclosing for the first time the technical details of the Seedream2.0 image generation model. This encompasses the entire process, from data construction and pre-training framework to post-training RLHF, marking a significant breakthrough in the text-to-image field. Since its launch on the Doubao app and Jimeng in early December 2024, Seedream2.0 has served over 100 million C-end users and has become a favorite among professional designers. Compared to mainstream models like Ideogram2.0 and Midjourney V6.1, it...

Against the backdrop of rapid advancements in artificial intelligence, research and applications of AI large models are booming. Recent developments from various companies and institutions in this field have garnered significant attention. Firstly, Nvidia's price-to-earnings ratio before the launch of ChatGPT attracted considerable attention on Wall Street, with experts suggesting this might foreshadow a repeat of Apple's success story in 2008. Meanwhile, Yu Feng, CIO of Guotai Junan Securities, stated that the rise of AI large models will propel the securities industry into an era of "intelligent cognition".

ByteDance's Doubao large model team recently announced a breakthrough in addressing key bottlenecks in Mixture-of-Experts (MoE) architecture, open-sourcing a significant optimization technology called COMET. This technology dramatically improves large model training efficiency, achieving a remarkable 1.7x speedup and a 40% reduction in training costs. Image Note: Image generated by AI, image licensing provider Midjourney. COMET has been deployed in ByteDance's multi-thousand-GPU cluster training, resulting in millions of GPU hours saved.

On March 10th, Zhiyuan Robotics officially launched its first general-purpose embodied base large model – Genie Operator-1 (GO-1). This announcement has garnered significant attention, particularly regarding its potential in home service robots and the new hope it offers for future household management. According to Zhiyuan Robotics' official introduction, the GO-1 large model, trained on a vast amount of human video data, demonstrates excellent performance in executing various household tasks such as delivering cups, preparing meals, and greeting guests. In terms of technical performance,

Recently, the Modelers community officially launched Step-Video and Step-Audio, two open-source multimodal large models developed by Step-Star. These models are designed for video generation and voice interaction, respectively, aiming to provide developers and enterprise users with more powerful AI tools. Step-Video, formally known as Step-Video-T2V, is a 30-billion parameter model, making it the world's largest open-source video generation model. This model can directly generate 20...

Reports indicate that the National Supercomputing Internet Platform has integrated Alibaba's Qwen large language model, officially providing the QwQ-32B API service. Users can access up to 1 million tokens free of charge, offering a valuable opportunity for developers and researchers. QwQ-32B is the latest open-source inference model from Alibaba's Tongyi team, demonstrating excellent performance. According to multiple authoritative benchmark tests, QwQ-32B's capabilities rival those of a fully-fledged 671B model.