Thinking Machines, the artificial intelligence startup founded by former OpenAI Chief Technology Officer Mira Murati, has officially launched its first AI model, Inkling, marking the company's e
ntry into the increasingly competitive generative AI market. The release is significant because Inkling is an open-weight model, meaning developers and organizations can access its model weights, customize its behavior, and fine-tune it for specialized tasks instead of relying solely on a closed, cloud-hosted service.
Inkling is built using a Mixture-of-Experts (MoE) architecture with 975 billion total parameters, although only 41 billion parameters are active during inference, making it more computationally efficient than a traditional dense model of the same size. The model supports an exceptionally large 1 million-token context window, enabling it to process long documents, extensive codebases, and complex research materials in a single conversation. It is also a multimodal AI model, capable of understanding text, images, and audio, making it suitable for a wide range of enterprise and developer applications.
One of Inkling's biggest differences from ChatGPT by OpenAI and Claude by Anthropic is its openness. ChatGPT and Claude are proprietary AI systems where users interact through hosted services, while their underlying model weights remain unavailable to the public. Inkling, on the other hand, allows organizations to download, self-host, modify, and fine-tune the model using their own datasets, making it particularly attractive for businesses with privacy, compliance, or customization requirements.
Thinking Machines has also introduced Tinker, its platform for model customization and fine-tuning. Instead of focusing solely on outperforming competitors in benchmark scores, the company says its goal is to provide a flexible foundation model that organizations can adapt for industry-specific tasks such as customer support, coding assistants, research, healthcare, finance, and enterprise automation. This customization-first approach sets Inkling apart from many commercial AI assistants that prioritize general-purpose performance.
Although Inkling performs competitively across reasoning, coding, multimodal understanding, and agent-based tasks, Thinking Machines acknowledges that it does not yet surpass the strongest frontier models from OpenAI, Anthropic, or Google in overall benchmark performance. Instead, the company positions Inkling as a practical, customizable AI platform that balances performance, cost, and developer flexibility.
The launch reflects a broader trend in the AI industry toward open and customizable models. As businesses increasingly seek greater control over how AI systems are deployed, Inkling could become an attractive alternative for organizations that want to run AI on their own infrastructure while maintaining the ability to tailor models for specific workflows. With this debut release, Thinking Machines has established itself as a serious new competitor in the rapidly evolving artificial intelligence landscape
