Mira Murati's Thinking Machines Lab releases 975B-parameter open-weight AI model
In brief
- Thinking Machines Lab, founded by OpenAI's former CTO Mira Murati in February 2025, launched Inkling on July 15 with 975 billion total parameters and 41 billion active parameters.
- Inkling is a multimodal model supporting text, audio, and video with a 1 million token context window, released under Apache 2.0 license for unrestricted use.
- Thinking Machines also released Inkling-Small (12B active parameters), available on Databricks and Hugging Face.
- Thinking Machines raised $2 billion in seed funding at a $10 billion valuation, positioning itself against Meta's Llama and Mistral.
- The startup's business model relies on enterprise support, customization services, and platform integrations rather than API gatekeeping.
Model Architecture and Capabilities
Inkling contains 975 billion total parameters in a Mixture-of-Experts architecture, with only 41 billion active at any given time. This efficiency-focused design reduces computational overhead while maintaining capability across domains. The model processes text, audio, and video and supports a context window of up to 1 million tokens, enabling long-form reasoning and multi-modal reasoning tasks.
Thinking Machines also released Inkling-Small, a lighter variant with 12 billion active parameters, targeting organizations with constrained compute budgets. Both models are available for immediate download and integrate with platforms like Databricks and Hugging Face, lowering barriers to deployment.
Open-Weight Strategy and Market Positioning
Murati is releasing Inkling under an Apache 2.0 license, meaning developers can download, modify, and deploy it with virtually no restrictions. This stands in sharp contrast to OpenAI and Anthropic's approach. OpenAI and Anthropic keep their most powerful models locked behind API access, controlling how customers interact with their systems and capturing all usage revenue.
Thinking Machines is betting on a different market. Enterprises want models they can customize, fine-tune, and run on their own infrastructure without handing their data to a third party. Meta has been the dominant player in open-weight AI with its Llama series, and Mistral has carved out a niche in Europe. Murati's entry raises the competitive bar considerably.
Funding and Viability Questions
Thinking Machines raised $2 billion in seed funding at a valuation of roughly $10 billion. That's a massive war chest for a startup founded just months after Murati's OpenAI departure. The capital signals investor confidence in the open-weight thesis, but it also creates a scaling challenge. Thinking Machines will need to prove that customization, enterprise support, and platform integrations can generate sufficient revenue to justify the $10 billion valuation.
There was no immediate market reaction to the Inkling launch, and Thinking Machines remains a private company. The real test will come as enterprises evaluate whether open-weight models can match the performance and reliability of closed alternatives while offering the flexibility and data privacy that on-premise deployment provides.
