AI startup Thinking Machines Lab, founded by former OpenAI CTO Mira Murati, has launched its first open-source AI model, Inkling. This model, a mixture-of-experts system, boasts 975 billion parameters but utilizes only 41 billion per task. Trained on 45 trillion multimodal tokens, Inkling is designed to provide calibrated responses and allows users to adjust processing speed by modifying the "thinking effort." Inkling is available for direct download and modification by developers and companies, offering a customizable solution for enterprises. It consumes only one-third of the tokens compared to Nvidia's Nemotron 3 Ultra for similar coding tasks. Thinking Machines aims to facilitate enterprise customization through its Tinker platform, positioning Inkling as a versatile tool for tailored AI solutions.