Thinking Machines Unveils Adaptable AI Model Inkling

Thinking Machines Lab, a company founded by former OpenAI CTO Mira Murati, has introduced Inkling, its first major AI model. The team designed it as an open foundation for developers who want to adapt artificial intelligence to their own tasks instead of relying only on the capabilities of ready-made closed systems.

Inkling is built on a Mixture-of-Experts Transformer architecture with 975 billion parameters, 41 billion of which are active during operation. The model supports a context window of up to one million tokens and can process text, images, audio, and video. Alongside the main model, Thinking Machines introduced Inkling-Small, a lighter version with 12 billion active parameters designed for scenarios where lower costs and reduced latency are important.

The team also focused on how the model performs in real-world applications. Inkling allows users to adjust its reasoning effort depending on the complexity of a task, helping balance response speed, resource consumption, and output quality. In addition, the model was trained to better evaluate the reliability of its own answers: when it lacks sufficient verified information, it is designed to acknowledge uncertainty instead of generating unsupported facts.

Inkling is already available for testing and fine-tuning through the Tinker platform, while its open weights allow developers to create customized versions of the model. The company has also made Inkling available through several AI infrastructure partners and says the release will serve as the foundation for a broader family of AI systems it plans to develop further.