Thinking Machines Lab, the artificial intelligence startup founded by former OpenAI chief technology officer Mira Murati, has released its first in-house AI model, Inkling, marking its first major public milestone after 18 months of largely behind-the-scenes infrastructure development.
Unlike the flagship models from OpenAI, Anthropic and Google, Inkling is open-weight, meaning outside developers and companies can download and modify it directly. The model uses a mixture-of-experts architecture with 975 billion total parameters, though it activates only about 41 billion for any given task — a design choice that keeps large models faster and cheaper to run.
For UK businesses, the implications are significant. The company is positioning Inkling less as a finished product and more as a starting point for organisations to fine-tune through its customisation platform, Tinker. This approach shifts responsibility for safety and performance onto the customer, which requires serious machine-learning expertise. It also echoes growing concerns from industry leaders: Microsoft CEO Satya Nadella argued last week that enterprises using proprietary AI models effectively pay twice — once in subscription fees and again by handing over business knowledge embedded in their prompts and corrections.
Thinking Machines does not claim Inkling is best-in-class, stating explicitly that it is 'not the strongest model available today, closed or open.' Instead, it aims for well-rounded performance and adaptability. On one benchmark, Inkling uses a third as many tokens as Nvidia's Nemotron 3 Ultra to achieve the same coding performance. The model also offers calibrated answers, flagging uncertainty rather than guessing, and lets users dial 'thinking effort' up or down to trade accuracy for speed.
The regulatory backdrop adds another layer for UK readers. The Information Commissioner's Office (ICO) has been scrutinising AI models for compliance with UK data protection law, while the EU AI Act imposes stricter rules on high-risk systems. Open-weight models like Inkling give organisations more control over data — potentially reducing compliance risks — but also place the onus on them to ensure safety and fairness. Hugging Face CEO Clem Delangue has predicted that frontier models will increasingly be reserved for experimentation, with most production AI work shifting to private or open-source alternatives — the exact split Thinking Machines is building around.