Thinking Machines launches the first open source large model Inkling, focusing on customization to combat "one-size-fits-all" AI. Thinking Machines Lab, an AI startup founded by former OpenAI Chief Technology Officer Mira Murati, officially released its first

Thinking Machines launches its first open source large model Inkling, focusing on customization to combat "one-size-fits-all" AI. Thinking Machines Lab, an AI startup founded by former OpenAI chief technology officer Mira Murati, officially released its first self-developed open source artificial intelligence model Inkling on July 15, 2026. The move marks the company’s first public showing of its sword after a year and a half of infrastructure construction, aiming to head-on challenge the “one-size-fits-all” closed-source AI business model dominated by mainstream technology giants by providing a highly customizable open weight model. Inkling adopts a hybrid expert (MoE) architecture and has 975 billion total parameters, but only needs to activate about 41 billion parameters in a single task, effectively balancing ultra-large scale and operational efficiency. The model is natively multi-modal pre-trained based on 45 trillion text, image, audio and video tokens and currently supports the output of text, code and structured data. Unlike general models that pursue "all-around", Inkling allows users to adjust "thinking intensity" according to business needs to balance speed and accuracy. In code benchmarks, it only requires one-third of the token consumption of NVIDIA Nemotron3 Ultra to achieve the same performance. As the core of its ecosystem, Thinking Machines positions Inkling as the starting point for enterprise-level fine-tuning, and cooperates with its model customization platform Tinker to realize monetization through account-sharing fine-tuning and hosting services. This strategy of avoiding the general chatbot path and focusing on enterprise privatization and customization is in line with the current industry trend of preventing proprietary data leaks and pursuing cost reduction and efficiency improvement. It also opens up a new path for the implementation of open source AI in enterprise-level production environments. via AI News (author: AI Base)