Peter H. Diamandis
July 25, 2026
Mira Murati's Contrarian AI Strategy: Customization Over Scale
TL;DR
Mira Murati's startup Thinking Machine Labs launched Inkling, an open-weight foundation model prioritizing customization and on-premise deployment over raw model power to compete in the AI race.
Key Takeaways
- 1Mira Murati, former OpenAI CTO, launched Inkling, an open-weight foundation model that can be downloaded, fine-tuned, and run on-premises on users' own hardware
- 2Inkling's strategy explicitly rejects leaderboard dominance, instead betting on AI companies' ability to customize the model for their own use cases
- 3The winning factor in AI competition is adaptability and customization leverage, not raw model power alone
- 4Open-weight models enable companies to maintain control over their AI systems rather than relying on frontier model providers for their data and workflows
Notable Quotes
“It's not going to be the future is the raw power. It's going to be the adaptability that's going to win.”
“She's built exactly the thing hitting the market that exactly what everybody needs right now.”
Chapters
1. Mira Murati's New Startup and Model Launch
Mira Murati, former CTO at OpenAI, launched Thinking Machine Labs with her first model called Inkling, an open-weight foundation model available for download and on-premises deployment
2. The Contrarian Strategy: Customization Over Scale
Instead of competing on leaderboards or raw power, Inkling's bet is on customization as the winning factor—allowing companies to adapt the model to their specific needs
3. Market Fit and Control
The model addresses immediate market demand by letting companies own their models, run them on-premises, and avoid surrendering control to large frontier model providers
Key People & Entities
- Mira Murati
- Former CTO at OpenAI, founder of Thinking Machine Labs
- Thinking Machine Labs
- Mira Murati's AI startup developing the Inkling model
- Inkling
- Open-weight foundation AI model by Thinking Machine Labs
Glossary
- Open-weight model
- An AI model whose weights (parameters) are publicly available and can be downloaded, modified, and run locally by users without restrictions
- Fine-tuning
- The process of adapting a pre-trained AI model to specific tasks or use cases by training it further on custom data
- On-premises (on-prem)
- Running software or models on a company's own servers and hardware rather than on cloud-based provider infrastructure
- Foundation model
- A large-scale AI model trained on broad data that can be adapted and fine-tuned for various downstream applications
- Frontier models
- State-of-the-art AI models developed by major providers that represent the cutting edge of AI capability