Thinking Machines Lab, led by former OpenAI CTO Mira Murati, this week released a fully open-source AI model called Inkling. It's a 975-billion-parameter behemoth, and it's the first model of that size to be released under a permissive Apache 2.0 license — meaning anyone can use, modify, and distribute it without royalties or restrictive clauses. The move has immediate implications for the crypto-AI sector, where open-weight models are prized for integration into decentralized applications and on-chain inference.
What makes Inkling different
Inkling isn't just big. It's the first fully open-source model at that scale — no gated access, no usage limits, no hidden layers. The lab released the full training code, architecture specs, and pretrained weights. That's a rarity in the world of large language models, where most big players release only partially open versions or restrict commercial use. The Apache 2.0 license means companies can fork Inkling, embed it into their products, and even sell access to it without worrying about legal friction.
Why crypto-AI is paying attention
The crypto-AI sector has been hungry for a model this open. Decentralized inference networks, token-gated AI agents, and on-chain data processors all need models that can be audited, fine-tuned, and deployed on commodity hardware. Inkling fits that bill. Its size — 975 billion parameters — puts it in the same league as GPT-4-class models, but without the proprietary strings. That could accelerate a wave of new crypto-native AI products, from autonomous trading bots to content generation protocols. The timing isn't an accident; Thinking Machines Lab has been signaling interest in the intersection of AI and decentralized computing for months.
Inside the Apache 2.0 choice
By choosing Apache 2.0, the lab sidestepped the more restrictive licenses often used by AI companies. That decision opens the door for crypto projects that need to redistribute models as part of their codebase. It also means developers can modify Inkling and keep their changes private — or contribute them back. The lab didn't put any usage caps or rate limits on the model itself, which is a break from the trend of API-gated access. For a model this size, it's a bet that wide adoption will create more value than walled-garden control.
Developers can already download Inkling from the lab's repository. The crypto-AI ecosystem is likely to start experimenting with it within days.
