Thinking Machines unveils open-weight AI model to rival Chinese open-source labs

Thinking Machines’ open‑weight launch is a reminder that the open model ecosystem is not just a US–China story. By positioning itself as a rare non‑Chinese alternative to the current crop of popular open-source models, the company is betting that governments and enterprises will want diversity in their base models for both geopolitical and technical reasons. Open‑weight, as opposed to fully open‑source, also reflects the post‑Mythos reality: labs want downstream customisation without conceding full control over training data and weights.

For the AGI race, the impact of any single open‑weight model is limited, but a robust competitive field outside China and the US matters. Open‑weight models become the substrate for regional fine‑tunes, sovereign AI projects and specialized vertical systems that might not justify licensing from a frontier lab. They also serve as a check on the narrative that only a handful of trillion‑parameter models matter.

If Thinking Machines can demonstrate credible performance while staying more permissive on use and modification than the biggest closed‑source players, it will add pressure on incumbents to clarify their own licensing and access tiers. Over time, that kind of competition at the mid‑frontier level could shape how much of the AGI stack ends up behind proprietary APIs versus embedded in customer‑controlled infrastructure.