Recently, the tech world and developer community have been captivated by an open-source project called Browser Use! This tool is like giving AI wings, allowing it to control browsers as naturally as humans. Using natural language, users can direct AI to automatically complete various web tasks. Its powerful automation capabilities and flexible deployment have ignited the passion of tech enthusiasts globally, creating a wave of excitement on X (formerly Twitter). Browser Use is rapidly advancing the field of AI-powered browser automation.

Cisco recently announced the formation of AGNTCY, a new open-source organization dedicated to providing critical infrastructure for the building and collaboration of AI agents. Cisco aims to unite AI and infrastructure experts to foster the development of an open and interoperable agent internet. Image note: Image generated by AI, licensed from Midjourney. With the official launch of AGNTCY, Cisco calls on experts to actively participate and contribute.

On March 6th, Open Source China (Open Source Consensus (Shanghai) Network Technology Co., Ltd.), a leading enterprise in the open-source technology ecosystem, announced the completion of hundreds of millions of yuan in Series C financing. This round of financing was led by Beijing Information Industry Development Investment Fund (Beijing Information Industry Fund), with Shenzhen Special Zone Daily Equity Investment Fund (Shenzhen Special Zone Daily) and Beijing Shanghe Momentum Private Equity Fund (Shanghe Momentum) following suit. Index Capital acted as the financial advisor. The funding will be used to deepen its AI strategy, expand its product matrix, promote intelligent solutions with software and hardware synergy, and facilitate the implementation of AI in industrial fields. Founder and Chairman

Alibaba has open-sourced a new inference large language model, QwQ-32B. Benchmarks show performance comparable to DeepSeek-R1, but with significantly reduced VRAM requirements.

Recently, Tongyi announced the open-sourcing of its latest Tongyi Wanxiang large model, Wan2.1. Wan2.1 is an AI model focused on high-quality video generation. With its excellent performance in handling complex movements, reproducing realistic physics, enhancing cinematic quality, and improving instruction following, it has become the preferred tool for creators, developers, and businesses embracing the AI era.

DeepSeek announced its second-day product of the open-source week: the first open-source EP communication library for MoE models, enabling full-stack optimization for mixed-expert model training and inference. DeepEP is a high-efficiency communication library specifically designed for Mixture-of-Experts (MoE) and Expert Parallelism (EP). It aims to provide high-throughput and low-latency many-to-many GPU kernel communication, commonly known as MoE routing and aggregation. DeepEP not only supports low-precision operations such as FP8 but also integrates with the DeepSeek-V3 paper.

On the first day of DeepSeek Open Source Week, we officially released our latest technological achievement, FlashMLA, an efficient Multi-Layer Attention decoding kernel specifically designed for NVIDIA Hopper architecture GPUs. This technology is optimized for variable-length sequence scenarios and can significantly enhance the inference performance of large models. The core technical features of FlashMLA include comprehensive support for BF16 precision and a paged key-value cache with a block size of 64.

Recently, Tencent's Hunyuan Video Model has officially begun recruiting testing partners on Platform X, marking a critical testing stage for this cutting-edge AI video generation technology. According to official sources, there is a high probability that the model will be open-sourced after the testing concludes, contributing its technological achievements to the global AI community. The Hunyuan Video Model is an important innovation by Tencent in the field of AI video generation, boasting over 13B parameters, making it one of the largest video generation models among open-source models.

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.