Inkling: Open Model from Thinking Machines Lab

Thinking Machines Lab, founded by former OpenAI CTO Mira Murati, has unveiled its first open-weight model — Inkling. It is a multimodal transformer using a Mixture-of-Experts architecture, with a total of 975 billion parameters, of which 41 billion are active. The model is trained on 45 trillion tokens, including text, images, audio, and video, and is distributed under the Apache-2.0 license. The company also announced a smaller version, Inkling-Small (276 billion parameters, 12 billion active), but its weights will be released later, after testing is complete. The developers emphasize that Inkling is not a frontier model, but is intended for fine-tuning using their Tinker platform. The model demonstrates competitiveness with open models from China and the United States, such as NVIDIA Nemotron and Gemma 4. The news also includes examples of API requests to the model for generating SVG and image descriptions.