Article · 2026-08-02 · 6 moments

Latest open artifacts (#23): Laguna S2.1, Inkling, & Kimi K3 show the utility of open models on the Pareto frontier

Capacity to train strong models is proliferating. ✦ AI generated

GUEST The author of 'Latest open artifacts (#23)...'HOST Nathan Lambert

Prediction

A safer bet than consolidation is predicting continued adoption of open models, and the ecosystem is entering a decisive era over how far open models can go.

Weighing the sustained pace of Chinese labs and new entrants like Xiaomi, the author now favors predicting continued adoption of open models and questions how far revenue-share licenses and open models can go.

transcript

The author of 'Latest open artifacts (#23)...': On the other side of the ecosystem is the sustained pace from the Chinese labs, with newer entrants like Xiaomi still accumulating mindshare in the broader AI economy. Having predicted consolidation for a long time, it now seems like a safer bet is to predict continued adoption, and try to imagine the role that open models play there. How much can revenue-share licenses like Kimi K3 stick? How much market share can open models take? We’re entering the decisive era.

supports · 5

Mechanism

Building token machines is a likely path to value, so labs that were expected to consolidate are instead finding value in continued operation.

With high demand for tokens expected to grow, labs realize that building token-production infrastructure is a viable value path, contrary to consolidation predictions.

transcript

The author of 'Latest open artifacts (#23)...': All of these labs we thought would need to consolidate are realizing that building token machines is a likely path to value, and more companies will identify that source of value over time.

gives example · 2supports · 1

Claim◆

Despite predictions of consolidation, more companies are training strong models and an increasing number of organizations are releasing these models openly.

The writer contrasts earlier predictions of lab consolidation with the current reality where more companies are investing hundreds of millions to billions and releasing open models.

transcript

The author of 'Latest open artifacts (#23)...': We’re at a place where more companies are training strong models — easily investing hundreds of millions to billions of dollars in the total effort still — and an increasing number of organizations are releasing these models openly.

explains mechanism · 1gives example · 1provides context · 1rebuts · 1supports · 1

Context

Kimi K3's noncommercial license requires inference and fine-tuning providers to enter commercial agreements, which could enable future US government action against US entities doing business with Chinese AI companies.

Kimi K3, the biggest open model release in some time, uses a noncommercial license; commentators argue this could expose US companies contracting with Moonshot to policy restrictions.

transcript

The author of 'Latest open artifacts (#23)...': It was released under a noncommercial license, requiring inference and fine-tuning providers to enter into a commercial agreement. Kevin Xu and Graham Webster argue in a post that these licenses enable potential future government action against US entities doing business with Chinese AI companies: 'But if a US company needs a contract with Moonshot to provide the inference tokens that Kimi K3 generates, the picture looks different. Some of the policy tools US officials and others have debated as potential levers to restrict Chinese open model use would more clearly apply.'

explains mechanism · 1

Fact

LongCat-2.0 is the first non-Huawei, non-toy model trained entirely on Chinese accelerators (Ascend 910s).

Meituan's LongCat-2.0, a 1.6T-parameter MoE, is not the most capable for its size but is notable as the first serious model trained entirely on Ascend 910 chips.

transcript

The author of 'Latest open artifacts (#23)...': While the model itself is not the most capable for its size beyond benchmarks, it was trained entirely on Ascend 910s, making it the first non-Huawei, non-toy model trained entirely on Chinese accelerators.

gives example · 1provides context · 1

Example◆

Thinking Machines has become a leading open-weight model company in the USA, with its open model finetuning service generating hundreds of millions in revenue per year.

The prime example against consolidation is Thinking Machines, which few predicted would go open but now earns hundreds of millions yearly from open model finetuning and releases the best open-weight models built in the USA.

transcript

The author of 'Latest open artifacts (#23)...': The prime example is Thinking Machines — when they announced their company in February 2025, very few people would’ve put them in the bucket of an open models company, myself included. Now their open model finetuning service is making hundreds of millions in revenue per year and they’re releasing the best open-weight models built in the U.S.A.

gives example · 1supports · 1

Highlight slides

The consolidation prediction that didn't materialize✦ from: Despite predictions of consolidation, more companies are training strong models and an increasing number of organizations are releasing these models openly.Thinking Machines: Leading US Open-Weight Model Company✦ from: Thinking Machines has become a leading open-weight model company in the USA, with its open model finetuning service generating hundreds of millions in revenue per year.Open release is growing✦ from: Despite predictions of consolidation, more companies are training strong models and an increasing number of organizations are releasing these models openly.From Surprise to Market Leader✦ from: Thinking Machines has become a leading open-weight model company in the USA, with its open model finetuning service generating hundreds of millions in revenue per year.

Related episodes

Audio · The Changelog

Community perspectives on Elastic vs AWS (Interview)

→

Video · All-In Podcast

The Fight Over Open Source AI, Anthropic's $1.5B Payout, NYC Socialists: Evictions = Violence?

→

Article · Big Technology

Everything You Need to Know About Kimi K3, the Latest Model From China to Shake Up AI

→