Thinking Machines Lab, founded by former OpenAI CTO Mira Murati, launched its first open-source AI model, Inkling, on July 15, 2026. The model uses a mixture of experts architecture with 975 billion total parameters but activates only 41 billion per task, balancing scale with efficiency. Pre-trained on 45 trillion multimodal tokens, Inkling allows users to adjust thinking intensity for speed or accuracy. It matches NVIDIA Nemotron3Ultra's code performance with one-third the token usage. The company avoids general chatbots, instead targeting enterprise customization through its Tinker platform and revenue-sharing fine-tuning, capitalizing on demand for secure, cost-efficient proprietary AI deployment.