41 Rivals Signed the Same Open-Weight Letter. That's Not Consensus, It's Nerves

Forty-one organizations, including direct rivals OpenAI and Meta, signed the same letter arguing that open-weight models are essential to American AI leadership. That kind of coalition is unusual enough on its own: cloud giants, startups, and competing labs rarely agree on policy language, let alone put their names on the same document. The letter's framing, that open-weight AI is America's best bet for maintaining its edge, reads less like confidence and more like a hedge against a trend the signatories can already see in the benchmark data.

The numbers behind the nerves

Four Chinese labs, Qwen, DeepSeek, MiniMax, and Kimi, have matched or beaten US frontier models on reasoning, coding, and document-parsing benchmarks, and they did it with genuine capability gains rather than just aggressive pricing. That distinction matters: a competitor winning on price is a market problem. A competitor winning on benchmarks while also shipping open weights is a strategic one, because it removes the two advantages the US labs have leaned on, superior capability and the leverage of keeping weights closed.

Kimi K3 illustrates the limit of that Chinese push, though: at 2.8 trillion parameters, it's the largest open model ever released, and it still doesn't beat the best proprietary systems overall. It wins on specific coding and agentic tasks, but the gap to frontier leaders like Claude Fable 5 and GPT-5.6 Sol remains real. Scale alone isn't closing the distance. That's a genuinely reassuring data point for the US labs, and it sits in tension with the anxious tone of the 41-company letter. If the picture were as dire as the coalition's messaging implies, Kimi K3 wouldn't still be trailing.

Draining moat, or just a smaller one?

Coverage framing this as a draining US moat captures the anxiety accurately but risks overstating the conclusion. What's actually happening is narrower and, in a sense, more interesting: the gap between the best proprietary model and the best open-weight model has shrunk, and the geographic center of open-weight leadership has shifted east. That's a real shift in leverage. It is not the same as the US losing its capability lead outright, which the Kimi K3 numbers still contradict.

What the 41-company letter actually reveals, read this way, is a coordination problem more than a capability crisis. US labs know that if the best open-weight models keep coming from Chinese labs, American developers, and eventually American infrastructure, will default to building on top of them, simply because they're free, capable, and available. The letter is an attempt to get ahead of that default by making the case for domestically-led open-weight development before the choice gets made for the industry by momentum alone. Signing a joint letter costs a company nothing and signals commitment to regulators cheaply. Actually releasing a frontier-competitive open-weight model, the kind that would settle the argument, is a different and much more expensive commitment, and notably, most of the 41 signatories haven't made it.