Gate News message, April 10: Tencent Hunyuan released the HY-Embodied-0.5 series embodied intelligence foundation models, and the 2B-parameter version has been officially open-sourced. The series models are designed for embodied intelligence agents in the real world, focusing on enhancing spatiotemporal perception and embodied reasoning capabilities, and can be applied to scenarios such as environmental prediction, interaction execution, and task planning.
The model family includes two versions: 2B and 32B. The 2B model is intended for edge devices and lightweight deployment, while the 32B model is used for complex reasoning tasks and higher-level intelligent decision-making. In terms of technical architecture, HY-Embodied-0.5 introduces a Mixture-of-Transformers (MoT) mechanism to implement multimodal partitioned computation, enhances perceptual representation capabilities through latent Tokens, and combines self-evolution after-training with policy-based distillation to achieve coordinated capability optimization across small and large models.
In terms of performance, the series models have performed exceptionally well across 22 benchmark tests. The 2B model surpasses same-scale SOTA models on 16 tasks, and the 32B version is already close to frontier levels, demonstrating strong embodied intelligence generalization capability and real-world application potential.
Disclaimer: The information on this page may come from third parties and does not represent the views or opinions of Gate. The content displayed on this page is for reference only and does not constitute any financial, investment, or legal advice. Gate does not guarantee the accuracy or completeness of the information and shall not be liable for any losses arising from the use of this information. Virtual asset investments carry high risks and are subject to significant price volatility. You may lose all of your invested principal. Please fully understand the relevant risks and make prudent decisions based on your own financial situation and risk tolerance. For details, please refer to
Disclaimer.
Related Articles
Ethereum Co-founder Lubin: AI Will Be Critical Turning Point for Crypto, But Tech Giant Monopoly Poses Systemic Risk
Ethereum co-founder Joseph Lubin emphasized the transformative potential of AI for the cryptocurrency sector while cautioning against the risks of centralization among tech giants. He envisions AI-driven autonomous transactions on blockchain and highlights the convergence of traditional finance with DeFi.
GateNews1h ago
Elon Musk Pushes 'Universal High Income' Checks as Ultimate Solution for AI Unemployment
Elon Musk advocates for a Universal High Income to combat AI-induced unemployment, envisioning a future with ample goods and zero inflation. In contrast, experts like Sam Altman raise concerns about job loss and propose protective measures for workers.
Coinpedia1h ago
DeepSeek Reportedly Launches First External Fundraising Round, Targets $10B+ Valuation and $300M+
DeepSeek, a Chinese AI startup, is negotiating its first external funding round, aiming for at least $300 million at a $10 billion valuation. Despite previous rejections of investment offers, its fundraising discussions are now reportedly underway.
GateNews1h ago
ChatGPT ads move into Australia and New Zealand: Free and Go users first, paid plans stay ad-free
OpenAI expanded ChatGPT advertising on April 17, 2023 to Australia, New Zealand, and Canada for Free and Go users, with no ads for paid users. This marks the second pathway toward AI commercialization and takes into account business and regulatory risks, where the presence of ads can promote paid conversions.
ChainNewsAbmedia3h ago
Hyundai Motor Group Reorganizes Around AI and Robotics, Targets 30,000 Atlas Robots by 2030
Hyundai Motor Group is restructuring to focus on AI and robotics, reducing traditional operations. It plans a $34.3 billion investment in robotics by 2030 and aims to launch a robotics-as-a-service model, collaborating with Google DeepMind and NVIDIA.
GateNews6h ago
NEA explores use of artificial intelligence in nuclear regulation
The NEA Working Group on New Technologies convened a workshop on March 25--26, focusing on how artificial intelligence can be applied to regulatory oversight and internal operations within nuclear authorities.
Summary
NEA workshop explored real-world AI applications in nuclear regulation, with
Cryptonews9h ago
