/data/photo/2025/08/04/68906d9a44e20.jpg)
Mira Murati Returns to AI Development with Inkling Model
Former OpenAI Chief Technology Officer Mira Murati has re-entered the artificial intelligence sector with the launch of Inkling, an open-weight AI model. Developed by her new startup, Thinking Machines Lab, the model features approximately 975 billion parameters and is designed for developers requiring high-performance, customizable systems. According to reports from Fortune, the model requires significant hardware memory, with a file size totaling 200 GB.
Inkling and the Shift to Open-Weight Models
The release of Inkling signals a strategic pivot toward open-weight architecture. Unlike proprietary, closed-source models that restrict user access to internal mechanics, open-weight models allow developers to download, inspect, and modify the system to fit specific use cases. Thinking Machines Lab intends for this flexibility to assist developers in building more tailored AI applications. While the startup has not yet released the full technical specifications, the move places Murati in direct competition with major industry players like OpenAI, Google, and Anthropic.
Did you know?
Open-weight models are distinct from fully “open source” models because while the weights—the learned parameters of the model—are made public, the training data and the specific training code often remain proprietary to the developer.
Thinking Machines Lab and Industry Competition
Murati founded Thinking Machines Lab after departing OpenAI in 2024. Her tenure at OpenAI saw her involved in the development of several high-profile generative AI products. The new venture has drawn industry attention due to its assembly of former researchers and executives from established AI firms. By entering the market with a large-scale, 975-billion-parameter model, Thinking Machines Lab is positioning itself to challenge the dominance of existing generative AI providers, including those from China and the United States.
Pro Tips for Developers
- Consider whether your project requires base model architecture or specialized domain knowledge.
Frequently Asked Questions
What is an open-weight AI model?
An open-weight model provides the public with the trained “weights” or parameters of the neural network, allowing users to run the model on their own hardware and perform fine-tuning, unlike closed models that only offer access via APIs.
How does Inkling compare to other models?
With 975 billion parameters, Inkling occupies a space among high-capacity models. Its primary differentiator is its open-weight status, which prioritizes user control and local customization over the service-based access models used by companies like OpenAI or Google.
Who is behind Thinking Machines Lab?
The startup was founded by Mira Murati, the former CTO of OpenAI, and includes a team of veteran researchers and executives from various large-scale technology and AI companies.
Are you interested in the technical evolution of open-source AI? Subscribe to our newsletter for the latest updates on model releases and industry shifts, or join the discussion in the comments section below.
Keep reading