Why Open-Weight AI Models Matter: Jensen Huang Backs Kimi K3

Jensen Huang, founder of Nvidia, posted his first tweet on X, sharing a jointly signed open‑weight AI letter that backs the open‑source release of the new Kimi K3 model. The letter, signed by more than 20 technology giants and investors—including Nvidia, Microsoft, Meta, Hugging Face, Mistral and Y Combinator—emphasizes the importance of open‑weight models for the future of AI.

The article recalls the 1980s open‑source software movement, which challenged the notion that only tightly controlled code could advance. That movement created a transparent ecosystem that now underpins most internet services and critical systems for large tech firms, the U.S. military, and federal agencies, providing more than just cost savings.

It argues that U.S. AI leadership depends not on a single frontier model but on a robust, open ecosystem that permeates every sector. Open‑weight models—models whose weights anyone can download, inspect, modify, and run on their own infrastructure—make advanced AI more accessible, flexible, and widely applicable, thereby expanding AI’s economic reach.

Open‑weight models lower entry barriers for startups, established companies, universities, and public institutions, allowing them to leverage cutting‑edge capabilities without training from scratch or paying high fees for each task. This enables organizations to match the right model to the right task at appropriate cost, reserving the most powerful models for truly frontier problems while running efficient specialized models elsewhere.

The article acknowledges unique risks: once released, weights leave the original developers’ control, making modified versions hard to track. However, the correct response is not to ban open weights but to use them to strengthen defenses. In an era where attackers exploit advanced AI, defenders need comparable models to detect, simulate, and mitigate emerging threats, and open models improve transparency and collective vulnerability remediation.

Open‑weight models also foster competition across model developers, cloud chips, applications, and services, driving innovation, reducing costs, and ensuring AI benefits are broadly shared. They give customers greater control over data and model behavior, allowing organizations to retain knowledge and value created with AI, which supports national sovereignty and prosperity.

Policy recommendations include expanding compute access for startups and researchers, investing in shared training resources (datasets, tools, evaluation frameworks), and avoiding premature restrictions that could stifle competition or push innovation abroad. Regulators should distinguish legitimate model‑distillation techniques—widely used for improvement and evaluation—from illegal extraction from closed models, addressing the latter with targeted legal and commercial frameworks.

Kimi K3, released on July 16, is highlighted as the largest open‑source model currently available (2.8 trillion parameters) with breakthrough architecture that approaches the programming capabilities of Fable 5. Demand surged, with its coding package and web interface sold out and throttled, and the full weights are slated for release by July 27. Communities worldwide have already produced deployment guides, and Hugging Face hosts a countdown page for the model.

Following Kimi K3’s launch, OpenAI and Anthropic accused Chinese firms of using model‑distillation to steal intellectual property, prompting U.S. Treasury discussions about restricting certain Chinese open‑source AI models. Approximately 200 startups and investors formed the “Little Tech Association,” warning that banning cheap open models would force them to pay high API fees to OpenAI and Anthropic, potentially crippling U.S. startups.

In conclusion, the article asserts that an open AI future can drive prosperity, enhance competition, reinforce U.S. technological leadership, reduce risks, and ensure AI’s benefits reach all layers of the economy, provided the right policy choices are made.