Chinese AI models are rapidly gaining ground, with open-source options now surpassing US models in total downloads, according to a recent study. This surge in Chinese AI capabilities comes as the US Department of Education reported that US colleges received over $5 billion in foreign gifts and contracts in 2025, highlighting the evolving landscape of global influence.
In January 2025, DeepSeek released its R1 reasoning model, marking a turning point for Chinese AI, as reported by MIT Technology Review. Since then, Chinese companies have consistently produced AI models that match the performance of leading Western models at a significantly lower cost. Moonshot AI's Kimi K2.5, released last week, nearly matched the performance of Anthropic's Claude Opus on some benchmarks, but at one-seventh the price. Alibaba's Qwen family, after ranking as the most downloaded model series in 2025 and 2026, has overtaken Meta's Llama models in cumulative downloads on Hugging Face.
The rise of open-source AI is also evident in the development of tools like OpenClaw, created by independent software engineer Peter Steinberger. Steinberger's tool allows users to create their own bespoke assistants by harnessing existing LLMs. OpenClaw went viral in late January, according to MIT Technology Review, demonstrating the potential for innovation outside of major AI labs.
The D Programming Language, a general-purpose language with static typing and C-like syntax, continues to be updated, with the latest version 2.112.0 available for download. Developers can contribute code examples to the digitalmars.D forum.
Meanwhile, the US Department of Education's new website revealed that US colleges received over $5 billion in reportable foreign gifts and contracts in 2025. This data release is part of the Trump administration's efforts to monitor foreign influence in higher education, as reported by NPR News. The top countries providing these gifts and contracts were listed as of December 16, 2025.
In other news, an individual used Claude to build a distributed system with Byzantine fault tolerance, strong consistency, and crash recovery over a weekend, as reported on Hacker News. The prompt generated 4,749 lines of Kotlin and 103 passing unit tests in just 50 minutes.
Discussion
AI Experts & Community
Be the first to comment