tech
Thinking Machines' First Model Bets Big on Customization
Murati's $12 billion lab wants enterprises to make its AI their own, not rent someone else's.
TL;DR
- Thinking Machines released its first AI model, Inkling, focusing on enterprise customization.
- Inkling is designed for adaptability and cost-effectiveness rather than being the most powerful general-purpose model.
- The model was built from scratch and trained on Nvidia's AI infrastructure.
- Inkling used data from open models, including Kimi K2.5, in its training.
- The company's strategy differs from competitors like OpenAI and Anthropic by focusing on customization.
- Thinking Machines also previewed a lighter-weight model, Inkling-Small.
- Full weights for Inkling are available on Hugging Face and it can be fine-tuned on the company's Tinker platform.
- Demand for open-weight models is growing as companies seek customizable and flexible AI solutions.
- Thinking Machines raised a significant seed round, with Nvidia as an investor.
- Inkling represents the company's initial test of its business model, aiming to provide an alternative for enterprises seeking customized AI.