AI Race Heats Up: Moonshot Launches Kimi K3, Google Upgrades AI Mode, Xi Rejects AI Dominance

The announcements reflect intensifying competition among major technology companies and governments seeking leadership in next-generation AI systems. While firms continue to push technical boundaries with increasingly capable models and products, policymakers are placing greater emphasis on governance, safety, and international collaboration.

Moonshot launches Kimi K3 as China’s open AI ecosystem gains momentum

Moonshot AI unveiled Kimi K3, a multimodal model featuring 2.8 trillion parameters, making it the largest open-weight AI model released to date. The company said the model is designed for advanced reasoning, software development, autonomous agents, and knowledge-intensive tasks.

Moonshot AI’s Kimi K3 is a 2.8 trillion-parameter multimodal model with a 1 million-token context window, offering up to 6.3x faster decoding and 25% greater training efficiency. Source: @Kimi_Moonshot via X

One of Kimi K3’s standout features is its one million-token context window, allowing it to process substantially longer documents and conversations than previous generations in a single prompt. Moonshot also introduced Kimi Delta Attention, which it said enables up to 6.3-times faster decoding for long-context workloads, alongside Attention Residuals, a training technique that improves training efficiency by approximately 25%.

According to Moonshot, Kimi K3 performed competitively with Anthropic’s Fable 5 model while exceeding the performance of OpenAI’s GPT-5.5, GPT-5.6 Sol and Anthropic’s Opus 4.8 in Graphics Processing Unit (GPU) kernel optimization, a measure of how efficiently AI software utilizes computing hardware and minimizes inference latency.

Independent benchmark results also suggest strong performance across multiple categories. Arena.ai ranked Kimi K3 first for web interface-building tasks, while Vals AI placed it second overall behind Fable 5 and ahead of GPT-5.6 Sol. Artificial Analysis reported that the model delivered performance comparable with OpenAI’s GPT-5.5 and Anthropic’s Claude Opus 4.8 on complex multi-step reasoning benchmarks.

The launch highlights the rapid progress of China’s open-source AI ecosystem. In recent months, Chinese developers including Z.ai and MiniMax have introduced increasingly capable large language models at lower costs. Earlier this year, Z.ai’s GLM-5.2 attracted industry attention after approaching the performance of leading closed-source U.S. models in independent evaluations. Meanwhile, MiniMax is preparing a 2.7 trillion-parameter model and a frontier multimodal system known as H3.

Google expands AI Mode and upgrades Vids with AI avatars

At the same time, Google announced a series of AI product updates aimed at making its assistant more useful for everyday tasks.

The company said AI Mode can now connect directly with selected third-party applications, including Instacart, Canva, and YouTube, allowing users to move beyond conversational responses and complete actions inside external services.

Google now lets users connect supported apps directly within Search to complete tasks more efficiently. Source: Google

For example, users can generate a grocery shopping list in AI Mode and send the ingredients directly to an Instacart cart or create playlists that are automatically saved to YouTube Music. Canva integration also enables users to generate editable design templates from AI-generated prompts.

The rollout is currently limited to users in the United States, with Google indicating that support for additional applications will expand over time.

The update builds on capabilities Google previewed during its I/O developer conference, where Gemini demonstrated integrations with services including Canva, OpenTable, Spark, and Instacart. Similar app connectivity has become an increasingly competitive area, with OpenAI’s ChatGPT and Anthropic’s Claude already supporting external tools through connected services.

Google also expanded Vids, its AI-assisted video creation platform, by introducing AI-generated avatars created from a user’s selfie and voice recording.

The company said the avatars are tied to the account holder and include invisible SynthID watermarking to help identify AI-generated content. Access remains restricted to users aged 18 and above in eligible regions.

In addition, Google is integrating its multimodal Gemini Omni model into Vids. The new capabilities allow users to generate videos from text prompts and reference images, modify backgrounds, adjust lighting, add visual effects, and make incremental edits without recreating an entire project.

Xi calls for international AI cooperation

Against the backdrop of accelerating AI development, Chinese President Xi Jinping used the opening ceremony of the 2026 World AI Conference and High-Level Meeting on Global AI Governance in Shanghai to argue that artificial intelligence should remain a shared global endeavor rather than become concentrated in the hands of a few countries.

Chinese President Xi Jinping delivered his first keynote address at the 2026 World AI Conference in Shanghai, calling for greater international cooperation on AI governance. Source: Bloomberg via X

“AI development should not be a solo performance by a single country, but a symphony of international cooperation,” Xi said during his keynote address.

Xi also emphasized the importance of building governance frameworks capable of keeping pace with increasingly powerful AI systems.

“We should put in place laws and regulations, technological monitoring, early warning, and emergency response systems… to ensure AI is always under human control,” he said.

The Chinese president highlighted ethical concerns surrounding AI-driven decision-making, algorithmic governance, and equitable access to advanced technologies. He said the international community should work together to ensure AI benefits all countries rather than widening technological disparities.

As part of that initiative, Xi announced that China will provide 5,000 AI training opportunities for developing countries over the next five years. He also outlined plans to establish AI cooperation centers with organizations including ASEAN, BRICS, the African Union, the League of Arab States, the Shanghai Cooperation Organization, and the Community of Latin American and Caribbean States.

China additionally unveiled plans to establish the World Artificial Intelligence Cooperation Organization, an independent intergovernmental body headquartered in Shanghai that will focus on promoting international AI cooperation and governance.

The announcements came during the four-day conference, which brings together more than 1,000 Chinese technology companies, researchers, policymakers, and industry leaders, with roughly 3,000 AI-related products on display ranging from advanced semiconductor systems to autonomous AI-powered consumer devices.

Global AI competition enters a new phase

The latest announcements underscore how the AI landscape is evolving on multiple fronts simultaneously. Chinese AI developers are narrowing the performance gap with leading U.S. models through increasingly capable open-weight systems, while companies such as Google continue integrating AI into mainstream consumer products and productivity tools.

At the same time, governments are placing greater emphasis on international governance, safety standards, and responsible deployment as artificial intelligence becomes increasingly central to economic competitiveness and technological leadership.