Search AI Products and News

At today's Baidu AI DAY technology open day, Baidu's commercial R&D team officially launched its self-developed video generation model MuseSteamer and the accompanying video product platform **HuiXiang**. This innovation aims to create a comprehensive video generation solution by combining generative AI and multimodal technology, to meet the strong demand for native content production in scenarios such as search, advertising, and recommendations. The MuseSteamer video generation model series is rich, currently including Turbo, Lite, Pro, and

A major breakthrough in 3D modeling technology! The PartCrafter project co-developed by Peking University, ByteDance, and Carnegie Mellon University has officially been unveiled. It can generate high-precision, structured 3D models from a single RGB image, completely overturning the traditional complex process of segmentation followed by reconstruction. This technology not only improves generation efficiency but also infers the 3D geometry of unseen structures, showcasing immense potential for AI in the 3D generation field. The AIbase editorial team has compiled the latest information to provide you with an in-depth analysis of PartCrafter.

Recently, Mistral AI, headquartered in Paris, France, in partnership with All Hands AI, has launched an open-source AI model designed specifically for software development — Devstral. This model boasts up to 24 billion parameters and is currently in the "research preview" stage, released under the Apache2.0 open license, allowing developers and enterprises to use it freely in commercial applications. The release of Devstral marks a significant step forward for AI in the programming field.

Google recently released a new version of its Gemma3 series, exciting many AI enthusiasts. Just a month after its initial launch, Google released a Quantization Aware Training (QAT) optimized version of Gemma3, aiming to significantly reduce memory requirements while maintaining model quality. Specifically, the QAT-optimized Gemma3 27B model reduces VRAM requirements from 54GB to 14.1GB, meaning users can now run it on a single NVIDIA RTX 3090.

Recently, an innovative technology named LHM (Large-scale Human Model for Animation) has achieved a significant breakthrough in the field of 3D human body reconstruction, bringing new development directions and application prospects to the field. Reconstructing animatable 3D human bodies from a single image has been a highly challenging task, plagued by ambiguities in geometry, appearance, and deformation separation. Current state-of-the-art research mostly focuses on static human modeling, and these methods often rely on synthetic 3D scans for training, which greatly limits their applications in...

The Lenovo ThinkSystem WA7785a G3 server has achieved a record-breaking throughput of 6708 tokens per second while running the 671B parameter DeepSeek large language model. This demonstrates the exceptional performance capabilities of the system for demanding AI workloads.

Recently, in Hangzhou, Zhejiang, the rise of the AI tool DeepSeek has seen increasing numbers of entrepreneurs leveraging its powerful capabilities for profit. News of a Gen Z streamer generating 330 million Yuan in sales in a single day using DeepSeek for a live stream has attracted significant attention. Reports indicate that many businesses and individuals have experimented with new sales methods with the help of DeepSeek. For example, a business owner from Yiwu successfully sold out their entire inventory using DeepSeek.

Wuhan recently unveiled policies to boost its AI industry, launching an "AI+" initiative this year to promote large model applications across more than 20 sectors. Based on the "Several Policy Measures to Promote the Development of the Artificial Intelligence Industry in Wuhan," the initiative comprises ten key measures, including support for technological breakthroughs, enhanced computing power, and model innovation, aiming to accelerate the construction of a full AI industry chain ecosystem. The policies will be application-scenario driven, focusing on core elements such as computing power, models, data, and talent.

Recently, an interdisciplinary research team from the Technical University of Munich, the Helmholtz Zentrum Munich, and ETH Zurich published an important study introducing an innovative framework called Moscot (Multi-Omics Single-Cell Optimal Transport), which successfully reconstructed the developmental trajectories of 1.7 million mouse embryonic cells at 20 time points. This study, published in the journal Nature, marks a significant breakthrough in the field of single-cell genomics. The design inspiration for the Moscot framework comes from the optimal transport theory of the 18th century.

Recently, the research team at Meta Reality Labs has jointly released an innovative generative model called 'Pippo', which can generate a dense turnaround video with a resolution of up to 1K from a randomly taken photo. This breakthrough technology marks another significant advancement in the fields of computer vision and image generation. The core of the Pippo model lies in the design of its multi-view diffusion transformer. Unlike traditional generative models, Pippo does not require any additional inputs, such as fitted parameters.

Recently, an unconventional AI duel has been taking place on the internet. From verbal skirmishes to chess matches, the competition between AIs not only showcases their language capabilities and thought processes but has also become a hot topic of discussion. A recent dialogue video between DeepSeek and GPT on Bilibili has garnered widespread attention, with over 3 million views and more than 500,000 interactions. In this duel, DeepSeek demonstrates exceptional linguistic expressiveness, employing rhetorical devices such as parallelism and metaphor, and even manages to deliver witty rhymes, showcasing its uniqueness.