Ali Baba’s Video Model Tops Charts, Figure Cuts Ties with OpenAI, Nearly 80,000 AI Layoffs Worldwide | Frontline News
As of April 11, 2026, the global competition in the AI and technology industry has entered a complex phase. From iterative advancements in foundational model technologies to breakthroughs in embodied intelligence, and an escalated arms race in computational power, the rivalry continues to intensify among the major camps of the US, Europe, and China. This article highlights the top ten key events in the global industry today, providing a perspective on the transformative changes underlying technological breakthroughs for decision-makers and industry managers.
Ali Baba’s HappyHorse-1.0 Claims Top Spot
On April 10, Ali Baba officially announced that its previously anonymous video model, HappyHorse-1.0, secured the top position in the Artificial Analysis Video Arena, a global AI video evaluation leaderboard. Developed by its ATH-AI innovation department, the model achieved an Elo score of 1379 in the text-to-video category, surpassing the second-place Seedance 2.0 by 106 points. In the image-to-video category, it set a new platform record with a score of 1411. This marks the first time a Chinese AI video model has led in a global authoritative blind test, breaking the longstanding dominance of Western companies in the generative AI visual sector.
Mistral Launches Voxtral Open Source TTS
French AI giant Mistral AI has launched the Voxtral TTS, an open-source text-to-speech model. It features an innovative hybrid architecture that combines autoregressive semantic token generation with flow-matched acoustic modeling, natively supporting nine major languages. The model boasts a first sound delay as low as 90 milliseconds and generates speech at six times the real-time speed. It can perform zero-shot voice cloning with just three seconds of reference audio, with a post-quantization memory footprint of only 3GB, making it deployable on edge devices. This reinforces Mistral’s position as a leader in the European open-source AI ecosystem.
Star Era Wins Global Firsts
In the field of embodied dexterous manipulation, the Benjie’s Humanoid Olympic Games announced that the Chinese company Star Era won three global firsts with its self-developed VLA embodied model, excelling in tasks such as peeling oranges, lock picking, and flipping socks. Notably, in two core tasks, Star Era achieved a record time of 49 seconds for hand-peeling without tools, improving the speed by 35% while reducing sample usage by 32%.
Zhiyuan Robotics’ GO-2 Model Accepted at CVPR 2026
On April 9, Zhiyuan Robotics unveiled its next-generation embodied foundation model, Genie Operator-2 (GO-2), which has been accepted for publication at the prestigious CVPR 2026 conference. This model introduces an innovative “action thinking chain” mechanism, bridging logical reasoning with precise action execution, thereby addressing longstanding industry challenges. In four authoritative benchmark tests, its average task success rate reached 98.5%, exceeding international mainstream models.
Stanford and MIT Release Meta-Harness Study
A joint team from Stanford and MIT has published a significant study on Meta-Harness, achieving automated end-to-end optimization of large model systems. The research demonstrates how the Harness framework, which guides data flow and tool scheduling, can be optimized using historical code retention and execution trajectories. Experiments show a 7.7% increase in text classification accuracy and a 75% reduction in context token consumption.
Figure AI Ends Collaboration with OpenAI
US humanoid robot startup Figure AI has confirmed the termination of its partnership with OpenAI, shifting focus to its self-developed humanoid robot AI brain, Helix. The collaboration began in 2024 and lasted just one year, as Figure AI aims to develop a competitive edge in real-time control for embodied intelligence.
Amazon AWS AI Services Annual Revenue Surpasses $15 Billion
Amazon CEO Andy Jassy revealed in the annual shareholder letter that AWS AI services have surpassed $15 billion in annual revenue, marking over 50% year-on-year growth. This revenue constitutes approximately 10% of AWS’s total annual revenue, highlighting the rapid expansion of enterprise AI services.
Anthropic Partners with Google and Broadcom
Anthropic has announced a long-term partnership with Google and Broadcom to develop a next-generation AI data center with a total power output of 3.5 gigawatts, expected to come online in phases starting in 2027. This collaboration represents Anthropic’s largest investment in computational power to date.
Global Tech Industry Layoffs Reach Nearly 80,000
According to recent statistics, the global tech industry has seen a total of 78,557 layoffs between January 1 and April 1, 2026, with over 50% attributed to AI and automation-related workforce reductions. This trend underscores the significant restructuring of employment in the tech sector due to AI innovations.
CoreWeave Secures $21 Billion AI Deal with Meta
CoreWeave has announced a $21 billion long-term partnership with Meta to support its AI infrastructure, bringing their total collaboration to $35.2 billion. This agreement is expected to enhance Meta’s AI development and deployment capabilities significantly.
Original article by NenPower, If reposted, please credit the source: https://nenpower.com/blog/global-ai-developments-alibabas-video-model-achieves-top-ranking-figure-splits-from-openai-and-nearly-80000-tech-layoffs-worldwide/
