Never seen anything like it.
Just now, NVIDIA founder Jensen Huang posted his first-ever tweet:
The account is newly registered, and the tweets seem rushed.
In his first post, Huang shared a letter signed by NVIDIA titled "Open Weights and U.S. AI Leadership," explaining why open-source models are important.
He stated that artificial intelligence will transform every industry, empower every company, and be built by every nation. Open-source models enhance security and cybersecurity, accelerate innovation and dissemination, and enable sovereignty. The world needs both cutting-edge proprietary models and cutting-edge open-source models.
Actually, this is an open letter signed by more than twenty technology giants, startups, and investment institutions, including NVIDIA, Microsoft, Meta, Hugging Face, Mistral, and YC.
They now appear to be joining forces to "compete" against the closed-source giants OpenAI and Anthropic.
Here is the full translation:
In the 1980s, early pioneers of open-source software challenged the prevailing notion that software could only advance under strict corporate control of its code. This movement fostered the creation of a transparent ecosystem where developers around the world could learn from, modify, and improve software. Today, open-source software developed by these communities underpins much of the internet and provides the foundational infrastructure for the world’s largest tech companies, the U.S. military, and federal agencies conducting scientific research, cybersecurity, and other critical missions. The significance of open source extends far beyond reducing software costs—it has established a shared foundation of knowledge upon which generations of American engineers and entrepreneurs have built their institutional autonomy.
Today, the United States faces a similar choice in the field of artificial intelligence. Evaluating U.S. leadership in AI does not depend on a single cutting-edge model, but rather on whether the U.S. can build a robust, open ecosystem and embed it across all sectors. This is crucial for creating opportunities for innovation and prosperity nationwide. It requires expanding access to AI, fostering competition, building a strong application layer, and giving Americans greater control over the technologies they rely on. Open-weight models—AI models that anyone can download, inspect, modify, and run on their own infrastructure—are a vital component of this foundation, as they make advanced AI more accessible, flexible, and widely applicable.
Open weights expand access to the AI economy. Startups, established companies, universities, and public institutions can leverage advanced models without having to train from scratch or pay premium prices for state-of-the-art models for every task. Open weights enable every organization to match the right model to the right task at the right cost, reserving frontier-scale capabilities for truly frontier problems and running efficient, specialized models everywhere else. This self-discipline will make AI economically sustainable, especially as its applications expand to billions of everyday tasks. For the U.S. to win the age of artificial intelligence, it must integrate AI into workflows in factories, hospitals, farms, classrooms, and neighborhood shops.
Open weights expand access to the AI economy. Startups, traditional businesses, universities, and public institutions can leverage advanced models without having to train from scratch or pay premium prices for state-of-the-art models for every task. Open weights enable every organization to match the right model to the right task at the right cost, reserving frontier-scale capabilities for truly frontier problems and running efficient, specialized models everywhere else. This discipline will make AI economically sustainable, especially as its applications expand to billions of everyday tasks. For the U.S. to win the age of artificial intelligence, it must embed AI into workflows in factories, hospitals, farms, classrooms, and neighborhood shops.
Open weights also foster competition, which is essential to ensuring that AI advancements are widely shared rather than concentrated in the hands of a few. By enabling numerous organizations to build, fine-tune, and deploy advanced models, open weights generate competition not only among model developers but also across cloud chips, applications, and services. This competition drives innovation, reduces costs, and extends the benefits of AI throughout the broader economy.
Open weights also give customers greater control. As organizations invest in AI, they want to ensure they are not locked into a single vendor and do not lose the knowledge and capabilities they build over time. Open weight models help provide this assurance by enabling organizations to maintain control over their data, evaluate models, customize them according to their needs, and deploy them wherever required by their business. Additionally, as organizations leverage AI to create value, open weights allow them to own that value through self-improving models, specialized expertise, and accumulated knowledge, thereby advancing U.S. sovereignty and prosperity.
Indeed, open weights do present certain unique risks. Once released, these weights are no longer under the control of their original developers, and modified versions can be difficult to track or reverse-engineer. However, the correct response to this risk is not to ban open weights. In an era where cyber attackers leverage advanced AI technologies, defenders need access to models with comparable capabilities to detect, simulate, and respond to emerging threats. Open models expand defensive capabilities, enhance transparency, and enable multiple teams to identify and patch vulnerabilities.
In fact, openness may be one of the most important pathways to achieving AI safety. Relying solely on closed models is not inherently secure: they can be compromised, misused, or fail in ways that external parties cannot detect. Concentrating advanced AI capabilities behind a small number of closed models exacerbates this risk, creating single points of failure, undermining competition, and placing critical technologies in the hands of only a few vendors. In contrast, open-weight models enable broad communities of researchers and developers to examine their behavior, identify vulnerabilities, develop safeguards, and continuously improve them over time. Just as open-source software has demonstrated that transparency is safer than obscurity, AI safety may depend on enabling more people to test and strengthen the models upon which society relies. It allows for rigorous benchmarking and evaluation, red teaming exercises, and protective measures grounded in actual harms—rather than assuming that closed systems are inherently safer.
A robust AI ecosystem is not guaranteed. Policymakers have a critical opportunity to act. This includes expanding access to computational resources for startups and researchers, investing in shared training resources (datasets, tools, evaluation frameworks), and preserving diversity in the frontier by avoiding premature restrictions on open models—restrictions that could stifle competition or drive innovation overseas. These measures must also focus on how to expand AI’s autonomous adoption across the entire economy through a strong application layer.
When building this ecosystem, policymakers should carefully distinguish between legitimate model development techniques and misappropriation. Model distillation—the process of using a model’s outputs to help train or improve another model—is a widely adopted technique for enhancing, evaluating, and validating models. It reflects a long-standing tradition of learning from, building upon, and improving existing technologies, a tradition that has driven innovation since the rise of the open-source software movement. In contrast, illicit extraction of value from proprietary models raises legitimate concerns. These concerns should be addressed through targeted legal and commercial frameworks, rather than through broad restrictions on technologies that play a vital role in AI innovation.
The age of artificial intelligence can be an era of prosperity. By making the right choices, open AI can expand opportunities, enhance competition, solidify America’s technological leadership, reduce risks, and ensure that the benefits of this extraordinary technology reach every level of our economy. Such a future is worth building—and America should lead the way.
Kimi K3, a sensation abroad
The obvious trigger for all of this is the latest large model, released and open-sourced on July 16. Kimi K3. Last week, Kimi The release of K3 has been a "nuclear-level" event in the tech community. It is not only the largest model currently accessible to the open-source community (with a total of 2.8 trillion parameters), but also introduces groundbreaking innovations in its underlying architecture, achieving programming capabilities close to those of Fable 5.
Currently, Kimi The encoding packages and web interface for K3 have both sold out and been rate-limited, demonstrating the intense enthusiasm within technical communities both domestically and internationally for the new model. Moonshot AI has stated that the full model weights for K3 will be made available by July 27. Hugging Face is now even... Kimi The new model sets up a countdown page:
The world is waiting. Major overseas tech communities have already begun circulating the "Deployment Readiness Guide" for K3.
However, Kimi After the release of K3, OpenAI and Anthropic felt threatened and began accusing Chinese companies of stealing intellectual property using model distillation technology, lobbying U.S. officials. It is reported that the U.S. Department of the Treasury and other agencies have begun considering measures such as banning certain Chinese open-source AI models.
This move also triggered panic among Silicon Valley startups. Nearly 200 companies and investment firms swiftly formed the “Little Tech Association,” warning the government that banning affordable open-source models would force startups to pay exorbitant API fees to OpenAI and Anthropic, directly jeopardizing numerous U.S. startups with limited funding.
Jensen Huang’s debut on X didn’t showcase GPUs or large models—it was a joint open letter strongly supporting open-weight AI, underscoring Silicon Valley leaders’ concerns over obstacles facing open-source models and pushing the debate between “open-source vs. closed-source” to a new peak.
How will this crackdown on open source play out?
Reference content:
https://www.microsoft.com/en-us/corporate-responsibility/topics/open-weight/
This article is from the WeChat public account "Machine Heart" (ID: almosthuman2014), authored by someone focused on open source.