CMSC stated on an interactive platform that it has deployed the DeepSeek series of models, currently primarily integrated into the company's office communication software to support employees' daily work needs. In the future, the company will comprehensively utilize R1-class strong reasoning models for business digitalization upgrades, exploring applications centered around core businesses such as intelligent investment advisory, intelligent investment, and intelligent investment research.

Nvidia has introduced Dynamo, a new software designed to significantly accelerate the performance of its DeepSeek AI technology, promising a 30-fold speed increase.

Kunlun Wanwei has open-sourced its Skywork R1V multimodal reasoning model, marking a significant advancement in Chinese AI capabilities. This powerful model represents a major leap forward in the field.

Kunlun Wanwei has officially released Skywork R1V (referred to as "R1V"), the world's first industrial-grade multimodal reasoning model. This 3.8 billion parameter model's performance is close to the well-known closed-source model DeepSeek-R1, and even surpasses it in several benchmark tests, outperforming a series of current state-of-the-art (SOTA) technologies. Kunlun Wanwei's decision to open-source R1V aims to promote technology sharing and progress, injecting new vitality into the global AI open-source community. R1V is distinguished by its superior multimodal reasoning capabilities.

Recently, the 360 ZhiNao team announced the successful replication of DeepSeek's reinforcement learning results and the official release of the open-source reasoning model Light-R1-14B-DS. This model surpasses DeepSeek-R1-Distill-Llama-70B and DeepSeek-R1-Distill-Qwen-32B in performance, becoming the industry's first 14B parameter model to achieve reinforcement learning effects. It significantly improves mathematical reasoning capabilities, outperforming most 32B-parameter models.

SiliconCloud officially announced that Trae, ByteDance's first AI-native Integrated Development Environment (IDE), has integrated with the SiliconCloud platform to provide developers with a more efficient programming experience. Trae now integrates multiple SiliconCloud models; users can access various coding models such as DeepSeek-R1, V3, QWQ-32B, and Qwen2.5-Coder simply by using an API key. This flexible selection caters to diverse development needs.

Li Auto announced the official launch of the web version of its AI assistant, Li Xiang Tongxue. Users can now access this intelligent assistant on their computers. This move signifies Li Auto's further expansion of its intelligent service ecosystem, achieving cross-scenario intelligent service coordination.

Silicon-Based Flow's official WeChat account announced that, effective immediately, the DeepSeek-R1 & V3 APIs on the SiliconCloud platform now support batch inference (BatchInference). Users can send requests to SiliconCloud via the batch API, unaffected by real-time inference rate limits, with task completion expected within 24 hours. Compared to real-time inference, DeepSeek-V3 batch inference pricing is reduced by 50%. From March 11th to March 18th, DeepSeek-R1 will enjoy a 75% price reduction.

SiliconCloud's platform now officially supports batch inference for DeepSeek-R1 and V3 APIs. Users can submit requests via batch APIs to process large datasets within 24 hours, overcoming the limitations of real-time inference rates. A key highlight is the significant price reduction; DeepSeek-V3 batch inference is substantially cheaper than real-time inference.

On March 11th, the Tongyi Lab team announced the open-sourcing of the R1-Omni model, marking a significant advancement in multimodal model development. This model integrates reinforcement learning with verifiable reward (RLVR) methods, focusing on improving reasoning capabilities and generalization performance in multimodal emotion recognition tasks. R1-Omni's training is divided into two stages. In the cold-start phase, the team fine-tuned the model using a combined dataset containing 580 video clips sourced from Explainable Multimodal Emotion...

Recently, realme UI product manager Kang Da Leo announced a significant update: the Xiaobu Assistant in realme UI 6.0 has officially integrated the full version of DeepSeek-R1. This update covers six realme models, including the realme GT7 Pro, realme GT7 Pro Racing Edition, realme Neo7, realme Neo7 SE, realme GT5 Pro, and realme GT6, bringing a more intelligent and efficient user experience to numerous realme phone users.

On March 10th, Chengpian, an application under Baidu Wenku, officially integrated the full version of DeepSeek-R1, providing users with more efficient and powerful AI assistance. Based on Baidu Wenku's massive content accumulation and AI capabilities, Chengpian, after this upgrade, can better meet users' diverse needs in learning, life, and creation.

On March 6, 2025, a new language model called **Light-R1-32B** was officially launched. Built upon the **Qwen2.5-32B-Instruct** model, this mathematical problem-solving tool, specially trained for exceptional mathematical abilities, low training costs, and reproducibility, is a significant highlight in the field of artificial intelligence. The development team xAI stated that Light-R1-32B not only surpasses similar models in performance but also provides significant benefits for academic research and practical applications.